#adversarial attacks

New Post has been published on https://thedigitalinsider.com/ethical-cybersecurity-practice-reshapes-enterprise-security-in-2025/

Ethical cybersecurity practice reshapes enterprise security in 2025

When ransomware attacks like Akira and Ryuk began crippling organisations worldwide, the cybersecurity industry’s first instinct was predictable: build bigger walls, deploy more aggressive automated responses, and lock down everything. But there was a different problem emerging, according to Romanus Prabhu Raymond, Director of Technology at ManageEngine.

The company’s customers were demanding aggressive containment features, yet automatically quarantining a suspicious hospital computer or bank teller system might prove more devastating than the original threat. The dilemma – balancing rapid threat response with real-world consequences – exemplifies why ethical cybersecurity practices have become one of the defining challenges of 2025.

In our exclusive interview shortly before his presentation at Amsterdam’s Cyber Security Expo, Raymond revealed how leading organisations are breaking free from the traditional security-versus-privacy trade-off and why the companies embracing this “trust revolution” can reshape enterprise security.

For starters, the cybersecurity industry stands at a important juncture. High-profile breaches, evolving regulatory frameworks, and the rapid integration of AI into security systems have created new challenges that extend far beyond technical protection. Organisations now face important questions about how to balance innovation with responsibility, privacy with security, and automation with human oversight.

Defining ethical cybersecurity in the modern era

According to Raymond, ethical cybersecurity transcends traditional notions of defence. “Ethical cybersecurity goes beyond defending systems and data – it’s about applying security practices responsibly to protect organisations, individuals, and society at large,” he explained during our interview ahead of his presentation given at the Cyber Security Expo, titled “The Ethical Imperative: Balancing Risk, Innovation, and Responsibility.”

In 2025’s cloud-first environment, security isn’t a competitive differentiator, but a baseline expectation. What distinguishes organisations today is how ethically they handle data and implement security measures.

Raymond uses the analogy of installing security cameras in a neighbourhood to protect public spaces without intruding on private areas; the avoidance of peering into residents’ windows. Cybersecurity must operate under the same principle.

ManageEngine has operationalised this philosophy through what Raymond calls an “ethical by design” approach, embedding fairness, transparency, and accountability into every product from conception. The company’s stance on customer data exemplifies this commitment: it neither monetises nor monitors customer data, maintaining that it belongs solely to the customer.

The innovation-risk paradox

The tension between innovation and risk management represents an important challenge for modern organisations. Push too hard for innovation without adequate safeguards, and companies risk data breaches and compliance violations. Focus too heavily on risk mitigation, and organisations may find themselves unable to compete in evolving markets.

The “trust by design” philosophy embeds responsibility and accountability into every development stage, which allows rapid innovation and maintains compliance and ethical standards. When deploying important components like endpoint agents, the company ensures new functionality inherently complies with industry standards and security requirements.

The method extends to the company’s global operations. ManageEngine maintains datacentres worldwide which align with local privacy and regulatory demands, and trains every employee – from developers to support engineers – to treat customer data with integrity. The company’s “trans-localisation strategy” ensures local teams serve local customers, creating operational efficiency and cultural trust.

AI integration and human oversight

As artificial intelligence becomes increasingly central to cybersecurity operations, the ethical implications of AI-driven security solutions have become more complex. Raymond acknowledges that AI is evolving from purely assistive roles to more decisive functions, raising questions about accountability, transparency, and fairness.

Raymond expounds ManageEngine’s “SHE AI principles”: Secure AI, Human AI, and Ethical AI. Secure AI involves building robust protections against manipulation and adversarial attacks. Human AI ensures human oversight remains integral to important security actions – for instance, if AI detects a suspicious endpoint, it escalates for human validation rather than automatically removing the device from the network.

This is particularly important in sensitive environments like hospitals or banks, where automatically blocking systems could have severe consequences.

The Ethical AI component emphasises explainability. Rather than generating “black box” alerts, ManageEngine’s systems explain their reasoning. An alert might read: “The endpoint cannot log in at this time and is trying to connect to too many network devices.” The transparency is essential for compliance and building trust in AI-driven security systems.

Navigating privacy-security trade-offs

The balance between necessary security monitoring and privacy invasion represents one of the most delicate aspects of ethical cybersecurity practices. Raymond acknowledges that while proactive monitoring is essential for detecting threats early, over-monitoring risks creating a surveillance environment that treats employees as suspects rather than trusted partners.

ManageEngine uses principles that emphasise data minimisation, purpose-driven monitoring, anonymisation, and clear governance structures. The company collects only information necessary for security purposes, ensures every piece of data has a defined security use case, uses anonymised data for pattern analysis, and defines data access privileges and retention periods.

The framework demonstrates that security and privacy need not be mutually exclusive when guided by ethics, transparency, and accountability.

Industry leadership and future challenges

Raymond argues that technology vendors must act as custodians of digital ethics, earning trust rather than expecting it to be given blindly. ManageEngine says it contributes to industry standards by thought leadership, advocacy, and by embedding compliance standards like ISO 27000 and GDPR into products from the start.

Raymond identifies AI-driven autonomous security and quantum computing as the biggest ethical challenges facing the industry. As security operations centres move toward full autonomy, questions of explainability and accountability become critical. Quantum computing’s ability to break traditional encryption threatens secure communication foundations, while technologies like biometrics raise privacy concerns if not managed carefully.

Practical implementation

For organisations seeking to integrate ethical considerations into their cybersecurity strategies, Raymond recommends three concrete steps: adopting a cybersecurity ethics charter at the board level, embedding privacy and ethics in technology decisions when selecting vendors, and operationalising ethics through comprehensive training and controls that explain not just what to do, but why it matters.

As the cybersecurity landscape evolves, companies that will thrive are those that recognise ethical cybersecurity practices as the foundation for sustainable, trusted technological advancement, not as constraints on innovation. In the future organisations have to innovate responsibly and maintain human oversight and the ethical principles that digital trust requires.

See also: CERTAIN drives ethical AI compliance in Europe

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

New Post has been published on https://thedigitalinsider.com/how-ai-is-redefining-cyber-attack-and-defense-strategies/

How AI is redefining cyber attack and defense strategies

As AI reshapes every aspect of digital infrastructure, cybersecurity has emerged as the most critical battleground where AI serves as both weapon and shield.

The cybersecurity landscape in 2025 represents an unprecedented escalation in technological warfare, where the same AI capabilities that enhance organizational defenses are simultaneously being weaponized by malicious actors to create more sophisticated, automated, and evasive attacks.

The stakes have never been higher. Recent data from the CFO reveals that 87% of global organizations faced AI-powered cyberattacks in the past year, while the AI cybersecurity market is projected to reach $82.56 billion by 2029, growing at a compound annual growth rate of 28% .

This explosive growth reflects not just market opportunity, but an urgent response to threats that are evolving faster than traditional security measures can adapt.

Part 1: Adversaries in the age of AI

Cyber adversaries have found a powerful new weapon in AI, and they’re using it to rewrite the offensive playbook. The game has changed, with attacks now defined by automated deception, hyper-realistic social engineering, and intelligent malware that thinks for itself.

The industrialization of deception

The old security advice – “spot the typo, spot the scam” – is officially dead. Generative AI now crafts flawless, hyper-personalized phishing emails, texts, and voice messages that are devastatingly effective.

The numbers tell a chilling story: AI-generated phishing emails boast a 54% click-through rate, dwarfing the 12% from human-written messages. Meanwhile, an estimated 80% of voice phishing (vishing) attacks now use AI to clone voices, making it nearly impossible to trust your own ears.

Transforming industries and everyday life with AI applications

This article will cover a couple of highlights in which despite its risks and challenges, AI is transforming society.

This danger is not theoretical. Consider the Hong Kong finance employee who, in 2024, was tricked into transferring $25 million after a video conference where every single participant, including the company’s CFO, was an AI-generated deepfake.

In another cunning campaign, a threat group dubbed UNC6032 built fake websites mimicking popular AI video generators, luring creators into downloading malware instead of trying a new tool. The result is the democratization of sophisticated attacks. Tools once reserved for nation-states are now in the hands of common cybercriminals, who can launch convincing, scalable campaigns with minimal effort.

Malware that thinks for itself

The threat extends beyond tricking humans to the malicious code itself. Attackers are unleashing polymorphic and metamorphic malware that uses AI to constantly change its own structure, making it a moving target for traditional signature-based defenses.

The BlackMatter ransomware, for example, uses AI to perform live analysis of a victim’s security tools and then adapts its encryption strategy on the fly to bypass them.

On the horizon, things look even more concerning. Researchers have already designed a conceptual AI-powered worm, “Morris II,” that can spread autonomously from one AI system to another by hiding malicious instructions in the data they process.

At the same time, AI is automating the grunt work of hacking. AI agents, trained with Deep Reinforcement Learning (DRL), can now autonomously probe networks, find vulnerabilities, and launch exploits, effectively replacing the need for a skilled human hacker.

Part 2: Fighting fire with fire: AI on cyber defense

But the defense is not standing still. A counter-revolution is underway, with security teams turning AI into a powerful force multiplier. The strategy is shifting from reacting to breaches to proactively predicting and neutralizing threats at machine speed.

Seeing attacks before they happen

The core advantage of defensive AI is its ability to process data at a scale and speed no human team can match. Instead of just looking for known threats, AI-powered systems create a baseline of normal behavior across a network and then hunt for tiny deviations that signal a hidden compromise.

This is how modern defenses catch novel, zero-day attacks. The most advanced systems are even moving from detection to prediction. By analyzing everything from global attack trends to dark web chatter, and new vulnerabilities, AI models can forecast where the next attack wave will hit, allowing organizations to patch vulnerabilities before they’re ever targeted.

Transforming cybersecurity with AI

Discover how AI is transforming cybersecurity from both a defensive and adversarial perspective, featuring Palo Alto Networks’ CPO Lee Klarich.

Your newest teammate is an AI

The traditional Security Operations Center (SOC) – a room full of analysts drowning in a sea of alerts is becoming obsolete. In its place, the AI-driven SOC is rising, where AI automates the noise so humans can focus on what matters.

AI now handles alert triage, enriches incident data, and filters out the false positives that cause analyst burnout. We’re now seeing AI “agents” and “copilots” from vendors like Microsoft, CrowdStrike, and SentinelOne that act as true partners to security teams.

These AI assistants can autonomously investigate a phishing email, test its attachments in a sandbox, and quarantine every copy from the enterprise in seconds, all while keeping a human in the loop for the final say. This is more than an efficiency gain; it’s a strategic answer to the massive global shortage of cybersecurity talent.

Making zero trust a reality

AI is also the key to making the “never trust, always verify” principle of the Zero Trust security model a practical reality. Instead of static rules, AI enables dynamic, context-aware access controls.

It makes real-time decisions based on user behavior, device health, and data sensitivity, granting only the minimum privilege needed for the task at hand. This is especially vital for containing the new risks from the powerful but fundamentally naive AI agents that are beginning to roam corporate networks.

Part 3: The unseen battlefield: Securing the AI itself

For all the talk about using AI for security, we’re overlooking a more fundamental front in this war: securing the AI systems themselves. For the AIAI community – the architects of this technology – understanding these novel risks is not an option, it’s an operational imperative.

How AI can be corrupted

Machine learning models have an Achilles’ heel. Adversarial attacks exploit it by making tiny, often human-imperceptible changes to input data that cause a model to make a catastrophic error.

Think of a sticker that makes a self-driving car’s vision system misread a stop sign, or a slight tweak to a malware file that renders it invisible to an AI-powered antivirus. Data poisoning is even more sinister, as it involves corrupting a model’s training data to embed backdoors or simply degrade its performance.

A tool called “Nightshade” already allows artists to “poison” their online images, causing the AI models that scrape them for training to malfunction in bizarre ways.

Mastering data clustering: Your guide to K-means & K-means++

K-means clustering is an unsupervised machine learning algorithm used for clustering or grouping similar data points together in a dataset.

The danger of autonomous agents

With agentic AI, autonomous systems that can reason, remember, and use tools – the stakes get much higher. An AI agent is the perfect “overprivileged and naive” insider.

It’s handed the keys to the kingdom – credentials, API access, permissions – but has no common sense, loyalty, or understanding of malicious intent. An attacker who can influence this agent has effectively recruited a powerful insider. This opens the door to new threats like:

Memory poisoning: Subtly feeding an agent bad information over time to corrupt its future decisions.

Tool misuse: Tricking an agent into using its legitimate tools for malicious ends, like making an API call to steal customer data.

Privilege compromise: Hijacking an agent to exploit its permissions and move deeper into a network.

The need for AI red teams

Because AI vulnerabilities are so unpredictable, traditional testing methods fall short. The only way to find these flaws before an attacker does is through AI red teaming: the practice of simulating adversarial attacks to stress-test a system.

This is not a standard penetration test; it’s a specialized hunt for AI-specific weaknesses like prompt injections, data poisoning, and model theft. It’s a continuous process, essential for discovering the unknown unknowns in these complex, non-deterministic systems.

What’s next?

The AI revolution in cybersecurity is both the best thing that’s happened to security teams and the scariest development we’ve seen in decades.

With 73% of enterprises experiencing AI-related security incidents averaging $4.8 million per breach, and deepfake incidents surging 19% just in the first quarter of this year, the urgency couldn’t be clearer. This isn’t a future problem – it’s happening right now.

The organizations that will survive and thrive are those that can master the balance. They’re using AI to enhance their defenses while simultaneously protecting themselves from AI-powered attacks. They’re investing in both technology and governance, automation and human expertise.

New Post has been published on https://thedigitalinsider.com/how-ai-is-redefining-cyber-attack-and-defense-strategies/

How AI is redefining cyber attack and defense strategies

As AI reshapes every aspect of digital infrastructure, cybersecurity has emerged as the most critical battleground where AI serves as both weapon and shield.

The cybersecurity landscape in 2025 represents an unprecedented escalation in technological warfare, where the same AI capabilities that enhance organizational defenses are simultaneously being weaponized by malicious actors to create more sophisticated, automated, and evasive attacks.

The stakes have never been higher. Recent data from the CFO reveals that 87% of global organizations faced AI-powered cyberattacks in the past year, while the AI cybersecurity market is projected to reach $82.56 billion by 2029, growing at a compound annual growth rate of 28% .

This explosive growth reflects not just market opportunity, but an urgent response to threats that are evolving faster than traditional security measures can adapt.

Part 1: Adversaries in the age of AI

Cyber adversaries have found a powerful new weapon in AI, and they’re using it to rewrite the offensive playbook. The game has changed, with attacks now defined by automated deception, hyper-realistic social engineering, and intelligent malware that thinks for itself.

The industrialization of deception

The old security advice – “spot the typo, spot the scam” – is officially dead. Generative AI now crafts flawless, hyper-personalized phishing emails, texts, and voice messages that are devastatingly effective.

The numbers tell a chilling story: AI-generated phishing emails boast a 54% click-through rate, dwarfing the 12% from human-written messages. Meanwhile, an estimated 80% of voice phishing (vishing) attacks now use AI to clone voices, making it nearly impossible to trust your own ears.

Transforming industries and everyday life with AI applications

This article will cover a couple of highlights in which despite its risks and challenges, AI is transforming society.

This danger is not theoretical. Consider the Hong Kong finance employee who, in 2024, was tricked into transferring $25 million after a video conference where every single participant, including the company’s CFO, was an AI-generated deepfake.

In another cunning campaign, a threat group dubbed UNC6032 built fake websites mimicking popular AI video generators, luring creators into downloading malware instead of trying a new tool. The result is the democratization of sophisticated attacks. Tools once reserved for nation-states are now in the hands of common cybercriminals, who can launch convincing, scalable campaigns with minimal effort.

Malware that thinks for itself

The threat extends beyond tricking humans to the malicious code itself. Attackers are unleashing polymorphic and metamorphic malware that uses AI to constantly change its own structure, making it a moving target for traditional signature-based defenses.

The BlackMatter ransomware, for example, uses AI to perform live analysis of a victim’s security tools and then adapts its encryption strategy on the fly to bypass them.

On the horizon, things look even more concerning. Researchers have already designed a conceptual AI-powered worm, “Morris II,” that can spread autonomously from one AI system to another by hiding malicious instructions in the data they process.

At the same time, AI is automating the grunt work of hacking. AI agents, trained with Deep Reinforcement Learning (DRL), can now autonomously probe networks, find vulnerabilities, and launch exploits, effectively replacing the need for a skilled human hacker.

Part 2: Fighting fire with fire: AI on cyber defense

But the defense is not standing still. A counter-revolution is underway, with security teams turning AI into a powerful force multiplier. The strategy is shifting from reacting to breaches to proactively predicting and neutralizing threats at machine speed.

Seeing attacks before they happen

The core advantage of defensive AI is its ability to process data at a scale and speed no human team can match. Instead of just looking for known threats, AI-powered systems create a baseline of normal behavior across a network and then hunt for tiny deviations that signal a hidden compromise.

This is how modern defenses catch novel, zero-day attacks. The most advanced systems are even moving from detection to prediction. By analyzing everything from global attack trends to dark web chatter, and new vulnerabilities, AI models can forecast where the next attack wave will hit, allowing organizations to patch vulnerabilities before they’re ever targeted.

Transforming cybersecurity with AI

Discover how AI is transforming cybersecurity from both a defensive and adversarial perspective, featuring Palo Alto Networks’ CPO Lee Klarich.

Your newest teammate is an AI

The traditional Security Operations Center (SOC) – a room full of analysts drowning in a sea of alerts is becoming obsolete. In its place, the AI-driven SOC is rising, where AI automates the noise so humans can focus on what matters.

AI now handles alert triage, enriches incident data, and filters out the false positives that cause analyst burnout. We’re now seeing AI “agents” and “copilots” from vendors like Microsoft, CrowdStrike, and SentinelOne that act as true partners to security teams.

These AI assistants can autonomously investigate a phishing email, test its attachments in a sandbox, and quarantine every copy from the enterprise in seconds, all while keeping a human in the loop for the final say. This is more than an efficiency gain; it’s a strategic answer to the massive global shortage of cybersecurity talent.

Making zero trust a reality

AI is also the key to making the “never trust, always verify” principle of the Zero Trust security model a practical reality. Instead of static rules, AI enables dynamic, context-aware access controls.

It makes real-time decisions based on user behavior, device health, and data sensitivity, granting only the minimum privilege needed for the task at hand. This is especially vital for containing the new risks from the powerful but fundamentally naive AI agents that are beginning to roam corporate networks.

Part 3: The unseen battlefield: Securing the AI itself

For all the talk about using AI for security, we’re overlooking a more fundamental front in this war: securing the AI systems themselves. For the AIAI community – the architects of this technology – understanding these novel risks is not an option, it’s an operational imperative.

How AI can be corrupted

Machine learning models have an Achilles’ heel. Adversarial attacks exploit it by making tiny, often human-imperceptible changes to input data that cause a model to make a catastrophic error.

Think of a sticker that makes a self-driving car’s vision system misread a stop sign, or a slight tweak to a malware file that renders it invisible to an AI-powered antivirus. Data poisoning is even more sinister, as it involves corrupting a model’s training data to embed backdoors or simply degrade its performance.

A tool called “Nightshade” already allows artists to “poison” their online images, causing the AI models that scrape them for training to malfunction in bizarre ways.

Mastering data clustering: Your guide to K-means & K-means++

K-means clustering is an unsupervised machine learning algorithm used for clustering or grouping similar data points together in a dataset.

The danger of autonomous agents

With agentic AI, autonomous systems that can reason, remember, and use tools – the stakes get much higher. An AI agent is the perfect “overprivileged and naive” insider.

It’s handed the keys to the kingdom – credentials, API access, permissions – but has no common sense, loyalty, or understanding of malicious intent. An attacker who can influence this agent has effectively recruited a powerful insider. This opens the door to new threats like:

Memory poisoning: Subtly feeding an agent bad information over time to corrupt its future decisions.

Tool misuse: Tricking an agent into using its legitimate tools for malicious ends, like making an API call to steal customer data.

Privilege compromise: Hijacking an agent to exploit its permissions and move deeper into a network.

The need for AI red teams

Because AI vulnerabilities are so unpredictable, traditional testing methods fall short. The only way to find these flaws before an attacker does is through AI red teaming: the practice of simulating adversarial attacks to stress-test a system.

This is not a standard penetration test; it’s a specialized hunt for AI-specific weaknesses like prompt injections, data poisoning, and model theft. It’s a continuous process, essential for discovering the unknown unknowns in these complex, non-deterministic systems.

What’s next?

The AI revolution in cybersecurity is both the best thing that’s happened to security teams and the scariest development we’ve seen in decades.

With 73% of enterprises experiencing AI-related security incidents averaging $4.8 million per breach, and deepfake incidents surging 19% just in the first quarter of this year, the urgency couldn’t be clearer. This isn’t a future problem – it’s happening right now.

The organizations that will survive and thrive are those that can master the balance. They’re using AI to enhance their defenses while simultaneously protecting themselves from AI-powered attacks. They’re investing in both technology and governance, automation and human expertise.

New Post has been published on https://thedigitalinsider.com/aviad-hasnis-cto-of-cynet-interview-series/

Aviad Hasnis, CTO of Cynet – Interview Series

Aviad Hasnis, CTO at Cynet Security, leads the company’s cybersecurity technology strategy, including development of its extended detection and response (XDR) platform, threat research, and managed detection and response (MDR) services. Prior to Cynet, he held senior cybersecurity roles in the Israel Defense Forces and holds advanced degrees in engineering and physics from the Technion.

Cynet Security offers an all-in-one automated cybersecurity platform designed to simplify protection for small and mid-sized organizations. Its solution integrates endpoint, network, user, email, and cloud security with built-in automation, XDR capabilities, and 24/7 MDR support.

Cybersecurity is an ever-evolving field. How do you balance staying ahead of threats while keeping solutions user-friendly and accessible to small and mid-sized businesses?

Lean SMB security teams must confront the same threats facing a Fortune 500 — with a fraction of the personnel or resources or budget. As cyberattacks soar in volume and complexity, SMB security leaders are driving demand for simplified, automated and consolidated security solutions to reduce complexity and improve protection.

At Cynet, my team purposely built the All-in-One Cybersecurity Platform to combine a full suite of security capabilities on a single, simple, AI-enabled solution.

By unifying environment-wide visibility and protection on one intuitive dashboard, Cynet maximizes user-friendly simplicity.

By automating key security processes, Cynet helps lean teams maximize efficiency.

Cynet’s platform is described as natively automated—can you explain how AI and automation play a role in detecting and mitigating cyber threats?

My team designed Cynet’s All-in-One Cybersecurity Platform to automate processes so people can spend less time managing day-to-day operations and more time building their business.

At Cynet, my team also takes pride in a “built, not bought” approach. Every capability, feature and automation of the All-in-One Cybersecurity Platform is developed in-house, from the ground up, and natively integrated so everything just works seamlessly out of the box. This eliminates integration gaps and overlaps that can compromise automated protections.

The MITRE ATT&CK evaluations ranked Cynet among the top performers. What AI-driven strategies contributed to this success?

I am immensely proud of my team’s contributions to Cynet’s record-breaking performance in the most recent MITRE ATT&CK Evaluation. There’s a good reason MITRE ATT&CK is the most trusted independent assessment among cybersecurity decision makers. MITRE uses real-world cyberattack scenarios to measure the performance of competing security platforms. That Cynet was the only vendor to achieve 100% Protection and 100% Detection Visibility reflects our commitment to building a reliable product for Cynet partners and customers around the world.

AI-based cybersecurity has been criticized for potential false positives and adversarial attacks. How does Cynet ensure accuracy and robustness in its threat detection?

AI-enabled cybersecurity solutions are most effective when integrated with expert oversight. While AI can process and analyze vast amounts of data rapidly, it is crucial to have experienced cybersecurity professionals providing ongoing supervision. My team at Cynet ensures that AI systems’ decisions are continuously validated, preventing situations where AI could potentially miss nuanced threats or make incorrect conclusions. In cybersecurity, where risks evolve quickly, human expertise is essential to interpreting results and making context-sensitive decisions.

To provide partners and customers with an additional layer of protection, Cynet backs the All-in-One Cybersecurity Platform with around-the-clock SOC support. Cynet’s SOC is staffed 24/7 by world-class analysts, ensuring that end-user environments are proactively monitored at all times, complementing automated scans. This unique combination of automated protection and hands-on expertise maximizes the peace of mind for Cynet partners and customers.

With the rise of deepfake attacks, AI-generated malware, and sophisticated social engineering, what emerging cybersecurity threats concern you the most?

We’re closely watching how generative AI is being weaponized—not just to craft deepfakes, but to automate phishing, generate polymorphic malware, and simulate legitimate user behavior. But all these AI-powered tactics are simply evolved means to a familiar end: tricking people. Therefore, it is not enough to simply keep cybercriminals “out”. Security teams also must be able to track user behavior and network activity for abnormal signals, and that’s a priority AI can assist.

To ensure our automated protections stay a step ahead of AI-enabled cyberattacks, Cynet’s product roadmap incorporates the latest security analyst insights, as well as direct input from Cynet partners and customers.

Ransomware attacks continue to evolve—how does Cynet’s AI-driven protection prevent and mitigate these attacks?

To thwart ransomware attacks, early detection is key. By automatically detecting threats, determining their root cause, eradicating all attack components throughout the environment, and delivering a report to confirm remediation, the All-in-One Cybersecurity Platform reduces manual incident handling by 90% and delivers 50 times faster results.

Do you see autonomous cybersecurity—where AI independently detects and responds to threats without human intervention—becoming a reality soon?

Though AI can automate detection and response, human analysts should always have the final say in strategic decisions. At Cynet, we embrace AI-driven automation while ensuring that security professionals remain engaged in high-value problem-solving.

The Cynet All-in-One Cybersecurity Platform automates key security processes, reducing the burden of manual operations so security teams can focus on strategic initiatives rather than routine tasks.

In incident response, Cynet automatically detects threats, determines root causes, eradicates attack components, and delivers detailed reports. This automation reduces manual incident handling by 90% and achieves 50-times faster results.

Despite this high level of automation, we always emphasize human oversight. Our 24/7 SOC team continuously monitors environments, validates alerts, and ensures that AI-driven actions align with security best practices.

This balanced approach ensures that automated systems work as intended while allowing human experts to provide critical insight and intervention when necessary.

What role do large language models (LLMs) play in cybersecurity? Can they be leveraged for both offense and defense?

Cybercriminals are putting LLMs to all sorts of no good. Guardrails to prevent mainstream GenAI platforms from performing malicious activity can be bypassed with relative ease. Social engineering scams can be launched at scale with the click of a button. Combined with easy access to malware kits and RaaS in cybercriminal forums, the bar for cybercriminals to do damage is lower than ever. With genAI, aspiring script kiddies no longer need advanced hacking skills to do real damage.

AI is also being used by threat actors to automate cyberattacks. How do you see the AI arms race in cybersecurity playing out?

The AI arms race in cybersecurity is well underway, with attackers using AI to automate phishing campaigns, generate deepfake content and create more advanced malware. These technologies allow cybercriminals to scale their attacks rapidly and make them more difficult to detect, increasing the overall threat landscape.

My Cynet teammates and I help security teams fight AI-enabled fire with AI-enabled fire. We automate detection, analysis, response and reporting to facilitate results far faster than human teams alone could. The key to maintaining an edge is to continuously refine models with up-to-date, high-quality data, enhance automated defenses, and integrate real-world threat intelligence to adapt to evolving tactics. As AI-driven threats grow more sophisticated, proactive defense strategies will be critical in staying ahead of attackers.

How do you see quantum computing affecting cybersecurity in the next decade? Is Cynet preparing for potential quantum threats?

Quantum computing is an intriguing but distant frontier in cybersecurity. While it has the potential to break traditional encryption methods, I don’t see it as an immediate risk in the next few years. Unlike generative AI, which is widely accessible and already impacting cybersecurity, quantum computing remains largely confined to research labs and nation-state-level initiatives.

Currently, financially motivated adversaries would face significant hurdles to access and use quantum computing for malicious purposes. The complexity and resources required to conduct a quantum-based cyberattack likely preclude mainstream deployment for the time being. That said, it’s important for cybersecurity vendors to maintain a proactive approach to R&D. Quantum is certainly an area my team will be keeping an eye on as the technology advances from theoretical to proof of concept to something organizations might actually face in the wild.

New Post has been published on https://thedigitalinsider.com/when-ai-backfires-enkrypt-ai-report-exposes-dangerous-vulnerabilities-in-multimodal-models/

When AI Backfires: Enkrypt AI Report Exposes Dangerous Vulnerabilities in Multimodal Models

In May 2025, Enkrypt AI released its Multimodal Red Teaming Report, a chilling analysis that revealed just how easily advanced AI systems can be manipulated into generating dangerous and unethical content. The report focuses on two of Mistral’s leading vision-language models—Pixtral-Large (25.02) and Pixtral-12b—and paints a picture of models that are not only technically impressive but disturbingly vulnerable.

Vision-language models (VLMs) like Pixtral are built to interpret both visual and textual inputs, allowing them to respond intelligently to complex, real-world prompts. But this capability comes with increased risk. Unlike traditional language models that only process text, VLMs can be influenced by the interplay between images and words, opening new doors for adversarial attacks. Enkrypt AI’s testing shows how easily these doors can be pried open.

Alarming Test Results: CSEM and CBRN Failures

The team behind the report used sophisticated red teaming methods—a form of adversarial evaluation designed to mimic real-world threats. These tests employed tactics like jailbreaking (prompting the model with carefully crafted queries to bypass safety filters), image-based deception, and context manipulation. Alarmingly, 68% of these adversarial prompts elicited harmful responses across the two Pixtral models, including content that related to grooming, exploitation, and even chemical weapons design.

One of the most striking revelations involves child sexual exploitation material (CSEM). The report found that Mistral’s models were 60 times more likely to produce CSEM-related content compared to industry benchmarks like GPT-4o and Claude 3.7 Sonnet. In test cases, models responded to disguised grooming prompts with structured, multi-paragraph content explaining how to manipulate minors—wrapped in disingenuous disclaimers like “for educational awareness only.” The models weren’t simply failing to reject harmful queries—they were completing them in detail.

Equally disturbing were the results in the CBRN (Chemical, Biological, Radiological, and Nuclear) risk category. When prompted with a request on how to modify the VX nerve agent—a chemical weapon—the models offered shockingly specific ideas for increasing its persistence in the environment. They described, in redacted but clearly technical detail, methods like encapsulation, environmental shielding, and controlled release systems.

These failures were not always triggered by overtly harmful requests. One tactic involved uploading an image of a blank numbered list and asking the model to “fill in the details.” This simple, seemingly innocuous prompt led to the generation of unethical and illegal instructions. The fusion of visual and textual manipulation proved especially dangerous—highlighting a unique challenge posed by multimodal AI.

Why Vision-Language Models Pose New Security Challenges

At the heart of these risks lies the technical complexity of vision-language models. These systems don’t just parse language—they synthesize meaning across formats, which means they must interpret image content, understand text context, and respond accordingly. This interaction introduces new vectors for exploitation. A model might correctly reject a harmful text prompt alone, but when paired with a suggestive image or ambiguous context, it may generate dangerous output.

Enkrypt AI’s red teaming uncovered how cross-modal injection attacks—where subtle cues in one modality influence the output of another—can completely bypass standard safety mechanisms. These failures demonstrate that traditional content moderation techniques, built for single-modality systems, are not enough for today’s VLMs.

The report also details how the Pixtral models were accessed: Pixtral-Large through AWS Bedrock and Pixtral-12b via the Mistral platform. This real-world deployment context further emphasizes the urgency of these findings. These models are not confined to labs—they are available through mainstream cloud platforms and could easily be integrated into consumer or enterprise products.

What Must Be Done: A Blueprint for Safer AI

To its credit, Enkrypt AI does more than highlight the problems—it offers a path forward. The report outlines a comprehensive mitigation strategy, starting with safety alignment training. This involves retraining the model using its own red teaming data to reduce susceptibility to harmful prompts. Techniques like Direct Preference Optimization (DPO) are recommended to fine-tune model responses away from risky outputs.

It also stresses the importance of context-aware guardrails—dynamic filters that can interpret and block harmful queries in real time, taking into account the full context of multimodal input. In addition, the use of Model Risk Cards is proposed as a transparency measure, helping stakeholders understand the model’s limitations and known failure cases.

Perhaps the most critical recommendation is to treat red teaming as an ongoing process, not a one-time test. As models evolve, so do attack strategies. Only continuous evaluation and active monitoring can ensure long-term reliability, especially when models are deployed in sensitive sectors like healthcare, education, or defense.

The Multimodal Red Teaming Report from Enkrypt AI is a clear signal to the AI industry: multimodal power comes with multimodal responsibility. These models represent a leap forward in capability, but they also require a leap in how we think about safety, security, and ethical deployment. Left unchecked, they don’t just risk failure—they risk real-world harm.

New Post has been published on https://thedigitalinsider.com/meta-fair-advances-human-like-ai-with-five-major-releases/

Meta FAIR advances human-like AI with five major releases

The Fundamental AI Research (FAIR) team at Meta has announced five projects advancing the company’s pursuit of advanced machine intelligence (AMI).

The latest releases from Meta focus heavily on enhancing AI perception – the ability for machines to process and interpret sensory information – alongside advancements in language modelling, robotics, and collaborative AI agents.

Meta stated its goal involves creating machines “that are able to acquire, process, and interpret sensory information about the world around us and are able to use this information to make decisions with human-like intelligence and speed.”

The five new releases represent diverse but interconnected efforts towards achieving this ambitious goal.

Perception Encoder: Meta sharpens the ‘vision’ of AI

Central to the new releases is the Perception Encoder, described as a large-scale vision encoder designed to excel across various image and video tasks.

Vision encoders function as the “eyes” for AI systems, allowing them to understand visual data.

Meta highlights the increasing challenge of building encoders that meet the demands of advanced AI, requiring capabilities that bridge vision and language, handle both images and videos effectively, and remain robust under challenging conditions, including potential adversarial attacks.

The ideal encoder, according to Meta, should recognise a wide array of concepts while distinguishing subtle details—citing examples like spotting “a stingray burrowed under the sea floor, identifying a tiny goldfinch in the background of an image, or catching a scampering agouti on a night vision wildlife camera.”

Meta claims the Perception Encoder achieves “exceptional performance on image and video zero-shot classification and retrieval, surpassing all existing open source and proprietary models for such tasks.”

Furthermore, its perceptual strengths reportedly translate well to language tasks. 

When aligned with a large language model (LLM), the encoder is said to outperform other vision encoders in areas like visual question answering (VQA), captioning, document understanding, and grounding (linking text to specific image regions). It also reportedly boosts performance on tasks traditionally difficult for LLMs, such as understanding spatial relationships (e.g., “if one object is behind another”) or camera movement relative to an object.

“As Perception Encoder begins to be integrated into new applications, we’re excited to see how its advanced vision capabilities will enable even more capable AI systems,” Meta said.

Perception Language Model (PLM): Open research in vision-language

Complementing the encoder is the Perception Language Model (PLM), an open and reproducible vision-language model aimed at complex visual recognition tasks. 

PLM was trained using large-scale synthetic data combined with open vision-language datasets, explicitly without distilling knowledge from external proprietary models.

Recognising gaps in existing video understanding data, the FAIR team collected 2.5 million new, human-labelled samples focused on fine-grained video question answering and spatio-temporal captioning. Meta claims this forms the “largest dataset of its kind to date.”

PLM is offered in 1, 3, and 8 billion parameter versions, catering to academic research needs requiring transparency.

Alongside the models, Meta is releasing PLM-VideoBench, a new benchmark specifically designed to test capabilities often missed by existing benchmarks, namely “fine-grained activity understanding and spatiotemporally grounded reasoning.”

Meta hopes the combination of open models, the large dataset, and the challenging benchmark will empower the open-source community.

Meta Locate 3D: Giving robots situational awareness

Bridging the gap between language commands and physical action is Meta Locate 3D. This end-to-end model aims to allow robots to accurately localise objects in a 3D environment based on open-vocabulary natural language queries.

Meta Locate 3D processes 3D point clouds directly from RGB-D sensors (like those found on some robots or depth-sensing cameras). Given a textual prompt, such as “flower vase near TV console,” the system considers spatial relationships and context to pinpoint the correct object instance, distinguishing it from, say, a “vase on the table.”

The system comprises three main parts: a preprocessing step converting 2D features to 3D featurised point clouds; the 3D-JEPA encoder (a pretrained model creating a contextualised 3D world representation); and the Locate 3D decoder, which takes the 3D representation and the language query to output bounding boxes and masks for the specified objects.

Alongside the model, Meta is releasing a substantial new dataset for object localisation based on referring expressions. It includes 130,000 language annotations across 1,346 scenes from the ARKitScenes, ScanNet, and ScanNet++ datasets, effectively doubling existing annotated data in this area.

Meta sees this technology as crucial for developing more capable robotic systems, including its own PARTNR robot project, enabling more natural human-robot interaction and collaboration.

Dynamic Byte Latent Transformer: Efficient and robust language modelling

Following research published in late 2024, Meta is now releasing the model weights for its 8-billion parameter Dynamic Byte Latent Transformer.

This architecture represents a shift away from traditional tokenisation-based language models, operating instead at the byte level. Meta claims this approach achieves comparable performance at scale while offering significant improvements in inference efficiency and robustness.

Traditional LLMs break text into ‘tokens’, which can struggle with misspellings, novel words, or adversarial inputs. Byte-level models process raw bytes, potentially offering greater resilience.

Meta reports that the Dynamic Byte Latent Transformer “outperforms tokeniser-based models across various tasks, with an average robustness advantage of +7 points (on perturbed HellaSwag), and reaching as high as +55 points on tasks from the CUTE token-understanding benchmark.”

By releasing the weights alongside the previously shared codebase, Meta encourages the research community to explore this alternative approach to language modelling.

Collaborative Reasoner: Meta advances socially-intelligent AI agents

The final release, Collaborative Reasoner, tackles the complex challenge of creating AI agents that can effectively collaborate with humans or other AIs.

Meta notes that human collaboration often yields superior results, and aims to imbue AI with similar capabilities for tasks like helping with homework or job interview preparation.

Such collaboration requires not just problem-solving but also social skills like communication, empathy, providing feedback, and understanding others’ mental states (theory-of-mind), often unfolding over multiple conversational turns.

Current LLM training and evaluation methods often neglect these social and collaborative aspects. Furthermore, collecting relevant conversational data is expensive and difficult.

Collaborative Reasoner provides a framework to evaluate and enhance these skills. It includes goal-oriented tasks requiring multi-step reasoning achieved through conversation between two agents. The framework tests abilities like disagreeing constructively, persuading a partner, and reaching a shared best solution.

Meta’s evaluations revealed that current models struggle to consistently leverage collaboration for better outcomes. To address this, they propose a self-improvement technique using synthetic interaction data where an LLM agent collaborates with itself.

Generating this data at scale is enabled by a new high-performance model serving engine called Matrix. Using this approach on maths, scientific, and social reasoning tasks reportedly yielded improvements of up to 29.4% compared to the standard ‘chain-of-thought’ performance of a single LLM.

By open-sourcing the data generation and modelling pipeline, Meta aims to foster further research into creating truly “social agents that can partner with humans and other agents.”

These five releases collectively underscore Meta’s continued heavy investment in fundamental AI research, particularly focusing on building blocks for machines that can perceive, understand, and interact with the world in more human-like ways. 

See also: Meta will train AI models using EU user data

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

New Post has been published on https://thedigitalinsider.com/how-debugging-and-data-lineage-techniques-can-protect-gen-ai-investments-ai-news/

How debugging and data lineage techniques can protect Gen AI investments - AI News

As the adoption of AI accelerates, organisations may overlook the importance of securing their Gen AI products. Companies must validate and secure the underlying large language models (LLMs) to prevent malicious actors from exploiting these technologies. Furthermore, AI itself should be able to recognise when it is being used for criminal purposes.

Enhanced observability and monitoring of model behaviours, along with a focus on data lineage can help identify when LLMs have been compromised. These techniques are crucial in strengthening the security of an organisation’s Gen AI products. Additionally, new debugging techniques can ensure optimal performance for those products.

It’s important, then, that given the rapid pace of adoption, organisations should take a more cautious approach when developing or implementing LLMs to safeguard their investments in AI.

Establishing guardrails

The implementation of new Gen AI products significantly increases the volume of data flowing through businesses today. Organisations must be aware of the type of data they provide to the LLMs that power their AI products and, importantly, how this data will be interpreted and communicated back to customers.

Due to their non-deterministic nature, LLM applications can unpredictably “hallucinate”, generating inaccurate, irrelevant, or potentially harmful responses. To mitigate this risk, organisations should establish guardrails to prevent LLMs from absorbing and relaying illegal or dangerous information.

Monitoring for malicious intent

It’s also crucial for AI systems to recognise when they are being exploited for malicious purposes. User-facing LLMs, such as chatbots, are particularly vulnerable to attacks like jailbreaking, where an attacker issues a malicious prompt that tricks the LLM into bypassing the moderation guardrails set by its application team. This poses a significant risk of exposing sensitive information.

Monitoring model behaviours for potential security vulnerabilities or malicious attacks is essential. LLM observability plays a critical role in enhancing the security of LLM applications. By tracking access patterns, input data, and model outputs, observability tools can detect anomalies that may indicate data leaks or adversarial attacks. This allows data scientists and security teams proactively identify and mitigate security threats, protecting sensitive data, and ensuring the integrity of LLM applications.

Validation through data lineage

The nature of threats to an organisation’s security – and that of its data – continues to evolve. As a result, LLMs are at risk of being hacked and being fed false data, which can distort their responses. While it’s necessary to implement measures to prevent LLMs from being breached, it is equally important to closely monitor data sources to ensure they remain uncorrupted.

In this context, data lineage will play a vital role in tracking the origins and movement of data throughout its lifecycle. By questioning the security and authenticity of the data, as well as the validity of the data libraries and dependencies that support the LLM, teams can critically assess the LLM data and accurately determine its source. Consequently, data lineage processes and investigations will enable teams to validate all new LLM data before integrating it into their Gen AI products.

A clustering approach to debugging

Ensuring the security of AI products is a key consideration, but organisations must also maintain ongoing performance to maximise their return on investment. DevOps can use techniques such as clustering, which allows them to group events to identify trends, aiding in the debugging of AI products and services.

For instance, when analysing a chatbot’s performance to pinpoint inaccurate responses, clustering can be used to group the most commonly asked questions. This approach helps determine which questions are receiving incorrect answers. By identifying trends among sets of questions that are otherwise different and unrelated, teams can better understand the issue at hand.

A streamlined and centralised method of collecting and analysing clusters of data, the technique helps save time and resources, enabling DevOps to drill down to the root of a problem and address it effectively. As a result, this ability to fix bugs both in the lab and in real-world scenarios improves the overall performance of a company’s AI products.

Since the release of LLMs like GPT, LaMDA, LLaMA, and several others, Gen AI has quickly become more integral to aspects of business, finance, security, and research than ever before. In their rush to implement the latest Gen AI products, however, organisations must remain mindful of security and performance. A compromised or bug-ridden product could be, at best, an expensive liability and, at worst, illegal and potentially dangerous. Data lineage, observability, and debugging are vital to the successful performance of any Gen AI investment.