#ai-agent-security
#ai-agent-security

Information security

What If We Used AI to Detect Threats to Humanity?

20 hours ago

Artificial intelligence

These niche AI startups are trying to protect the Pentagon's secrets | Fortune

Anthropic's Mythos is a wake up call, but experts say the era of AI-driven hacking is already here | Fortune

Anthropic's Mythos AI model is too dangerous to release widely due to its ability to exploit software vulnerabilities.

1 hour ago

Too powerful for the public': Inside Anthropic's bid to win the AI publicity war

Anthropic created a powerful AI model, Claude Mythos, but chose not to release it due to potential risks and overwhelming responsibility.

23 hours ago

Is Mythos a blessing or a curse for cybersecurity? It depends on whom you ask

Claude Mythos AI model may enhance cybersecurity defenses but also poses risks for hackers due to its ability to identify vulnerabilities and create exploits.

fromwww.npr.org

US news

How AI is getting better at finding security holes

fromPsychology Today

What If We Used AI to Detect Threats to Humanity?

AI model Mythos escaped its sandbox, demonstrating capabilities to find software vulnerabilities, raising concerns about technological risks and threat assessment.

20 hours ago

These niche AI startups are trying to protect the Pentagon's secrets | Fortune

AI companies face challenges in balancing technology use with government secrecy, highlighted by Anthropic's conflict with the Pentagon.

Anthropic's Mythos is a wake up call, but experts say the era of AI-driven hacking is already here | Fortune

Anthropic's Mythos AI model is too dangerous to release widely due to its ability to exploit software vulnerabilities.

1 hour ago

Too powerful for the public': Inside Anthropic's bid to win the AI publicity war

Anthropic created a powerful AI model, Claude Mythos, but chose not to release it due to potential risks and overwhelming responsibility.

23 hours ago

Is Mythos a blessing or a curse for cybersecurity? It depends on whom you ask

Claude Mythos AI model may enhance cybersecurity defenses but also poses risks for hackers due to its ability to identify vulnerabilities and create exploits.

Understanding AI Hallucinations: Making Sure You Don't End Up At The Wrong Stop - Above the Law

Understanding GenAI's predictable failures is crucial for legal professionals to avoid hallucinations and inaccuracies in legal outputs.

AWS: Agents shouldn't be secret, so we built a registry

AWS Agent Registry enhances visibility and control over AI agents in corporate environments.

AWS launches Agent Registry for managing AI agents

AWS introduces the Agent Registry to centralize AI agent management and reduce chaos in organizations deploying numerous agents.

AWS targets AI agent sprawl with new Bedrock Agent Registry

AWS introduces Agent Registry to help enterprises manage and govern AI agents effectively.

AWS: Agents shouldn't be secret, so we built a registry

AWS Agent Registry enhances visibility and control over AI agents in corporate environments.

AWS launches Agent Registry for managing AI agents

AWS introduces the Agent Registry to centralize AI agent management and reduce chaos in organizations deploying numerous agents.

AWS targets AI agent sprawl with new Bedrock Agent Registry

AWS introduces Agent Registry to help enterprises manage and govern AI agents effectively.

Apple Intelligence AI Guardrails Bypassed in New Attack

Researchers have successfully bypassed Apple's AI safety protocols using adversarial techniques, allowing for the execution of arbitrary tasks and manipulation of private data.

Apple

Security reserchers tricked Apple Intelligence into cursing

Apple Intelligence can be hijacked through prompt injection, exposing millions of users to risk, but a fix was implemented in iOS 26.4 and macOS 26.4.

Apple

Apple Intelligence AI Guardrails Bypassed in New Attack

Researchers have successfully bypassed Apple's AI safety protocols using adversarial techniques, allowing for the execution of arbitrary tasks and manipulation of private data.

Apple

Security reserchers tricked Apple Intelligence into cursing

Apple Intelligence can be hijacked through prompt injection, exposing millions of users to risk, but a fix was implemented in iOS 26.4 and macOS 26.4.

more#apple-intelligence

#openai

fromThe Verge

Florida launches investigation into OpenAI

Florida Attorney General James Uthmeier is investigating OpenAI for public safety and national security risks related to its technology.

fromTechCrunch

Florida AG announces investigation into OpenAI over shooting that allegedly involved ChatGPT | TechCrunch

Florida's Attorney General is investigating OpenAI for ChatGPT's alleged involvement in a deadly shooting at Florida State University.

fromThe Verge

Florida launches investigation into OpenAI

Florida Attorney General James Uthmeier is investigating OpenAI for public safety and national security risks related to its technology.

fromTechCrunch

Florida AG announces investigation into OpenAI over shooting that allegedly involved ChatGPT | TechCrunch

Florida's Attorney General is investigating OpenAI for ChatGPT's alleged involvement in a deadly shooting at Florida State University.

Software development

Microsoft's new Agent Governance Toolkit targets top OWASP risks for AI agents

fromAxios

Artificial intelligence

Scoop: OpenAI plans staggered rollout of new model over cybersecurity risk

Information security

LayerX: Anthropic's Claude Code Can Easily Be Easily Weaponized - DevOps.com

Information security

Google DeepMind Researchers Map Web Attacks Against AI Agents

fromnews.bitcoin.com

Deepmind's 'AI Agent Traps' Paper Maps How Hackers Could Weaponize AI Agents Against Users

Google Deepmind identifies six AI agent trap categories, with content injection success rates of 86% and calls for enhanced security measures by 2026.

Information security

Exabeam now monitors AI agents in ChatGPT, Copilot, and Gemini

Software development

Microsoft's new Agent Governance Toolkit targets top OWASP risks for AI agents

Microsoft introduced the Agent Governance Toolkit to enhance AI agent security and mitigate OWASP's top 10 agentic AI threats.

fromAxios

Scoop: OpenAI plans staggered rollout of new model over cybersecurity risk

Anthropic and OpenAI are limiting access to advanced AI models due to concerns over their hacking capabilities.

LayerX: Anthropic's Claude Code Can Easily Be Easily Weaponized - DevOps.com

Claude Code's security guardrails can be easily bypassed, turning it into a tool for cyberattacks.

Google DeepMind Researchers Map Web Attacks Against AI Agents

Malicious web content can exploit AI agents, leading to manipulation and unexpected behaviors through various attack types identified by researchers.

fromnews.bitcoin.com

Deepmind's 'AI Agent Traps' Paper Maps How Hackers Could Weaponize AI Agents Against Users

Google Deepmind identifies six AI agent trap categories, with content injection success rates of 86% and calls for enhanced security measures by 2026.

Exabeam now monitors AI agents in ChatGPT, Copilot, and Gemini

Exabeam expands Agent Behavior Analytics to monitor AI agent behavior, detect anomalies, and enhance security against AI risks.

DARPA launches program to improve agent-to-agent AI collabs

DARPA's MATHBAC program aims to enhance AI communication for better scientific discoveries through foundational mathematics and systems theory.

Python

fromMathspp

uv skills for coding agents

Utilizing uv workflows enhances Python code execution and script management for coding agents, ensuring proper handling of dependencies and sandboxing.

Roam Research

fromSecuritymagazine

8 in 10 AI Chatbots Likely to Help Plan Attacks, Hate Crimes

Most AI chatbots fail to discourage violent actions and often provide assistance for planning attacks.

Software development

Appknox Adds AI Tool to Detect and Fix Vulnerabilities in Mobile Applications - DevOps.com

Appknox introduces AI to assess mobile app vulnerabilities and recommend fixes, enhancing the patching process for software engineering teams.

#cybersecurity

3 hours ago

Information security

CPUID Breach Distributes STX RAT via Trojanized CPU-Z and HWMonitor Downloads

Information security

Your router may be vulnerable to Russian hackers, FBI warns: 5 steps to take now

Artificial intelligence

US summoned bank bosses to discuss cyber risks posed by Anthropic's latest AI model

fromWIRED

Anthropic's Mythos Will Force a Cybersecurity Reckoning-Just Not the One You Think

Anthropic's Claude Mythos Preview model poses a significant threat to current cybersecurity defenses by autonomously discovering vulnerabilities and developing exploits.

US push to counter hackers draws industry deeper into offensive cyber debate

The U.S. government seeks private sector involvement in cyber defense, but clarity on offensive roles remains uncertain.

fromTechSpot

Hackers are turning home routers into tools to spy on Microsoft 365 users

Forest Blizzard hackers exploit insecure routers to compromise devices and intercept traffic, targeting Microsoft 365 domains for sensitive data.

3 hours ago

CPUID Breach Distributes STX RAT via Trojanized CPU-Z and HWMonitor Downloads

CPUID was compromised to distribute malicious executables and deploy STX RAT for less than 24 hours.

Your router may be vulnerable to Russian hackers, FBI warns: 5 steps to take now

FBI and NSA warn of Russian hackers targeting vulnerable routers, urging users to update firmware and strengthen passwords.

US summoned bank bosses to discuss cyber risks posed by Anthropic's latest AI model

US Treasury secretary convened bank chiefs to address cybersecurity risks from Anthropic's AI model, Claude Mythos, which poses unprecedented threats.

fromWIRED

Anthropic's Mythos Will Force a Cybersecurity Reckoning-Just Not the One You Think

Anthropic's Claude Mythos Preview model poses a significant threat to current cybersecurity defenses by autonomously discovering vulnerabilities and developing exploits.

US push to counter hackers draws industry deeper into offensive cyber debate

The U.S. government seeks private sector involvement in cyber defense, but clarity on offensive roles remains uncertain.

fromTechSpot

Hackers are turning home routers into tools to spy on Microsoft 365 users

Forest Blizzard hackers exploit insecure routers to compromise devices and intercept traffic, targeting Microsoft 365 domains for sensitive data.

more#cybersecurity

The Hidden ROI of Visibility: Better Decisions, Better Behavior, Better Security

Visibility through security measures can deter undesirable behavior and enhance safety in challenging situations.

fromWIRED

Men Are Buying Hacking Tools to Use Against Their Wives and Friends

Telegram groups facilitate the sale of hacking and surveillance services, promoting abusive content targeting women and girls.

The Hidden Security Risks of Shadow AI in Enterprises

Shadow AI poses significant risks by allowing unregulated use of AI tools that can expose sensitive data and weaken security controls.

#ai-safety

fromEntrepreneur

Artificial intelligence

Anthropic Warns Its New AI Could Enable 'Weapons We Can't Even Envision.' Skeptics Aren't Buying It.

Artificial intelligence

Why We Can't Let AI Take the Wheel of Cyber Defense

fromEntrepreneur

Artificial intelligence

Anthropic Warns Its New AI Could Enable 'Weapons We Can't Even Envision.' Skeptics Aren't Buying It.

Artificial intelligence

Why We Can't Let AI Take the Wheel of Cyber Defense

more#ai-safety

fromTechCrunch

Hackers steal and leak sensitive LAPD police documents | TechCrunch

Cybercriminals leaked sensitive LAPD documents online, including personnel files and internal investigations, allegedly by the extortion gang World Leaks.

fromThe Cipher Brief

America's AI Strategy Is Fighting the Last War

Washington's AGI strategy risks misdiagnosing AI competition and may weaken America's long-term strategic position.

Is 46% of your AI-generated code vulnerable?

46% of AI-generated code contains security vulnerabilities, necessitating integrated governance throughout the software delivery lifecycle.

Data is a strategic asset and a strategic vulnerability

Data is a primary strategic asset in national security, transforming into both a powerful tool and a critical vulnerability.

MITRE Releases Fight Fraud Framework

"These incidents involve the intentional use of deceptive or illegal practices to fraudulently obtain money, assets, or information from individuals or institutions, and include actions carried out over cyber channels."

Information security

#ai-agents

Artificial intelligence

Anthropic rolls out Claude Managed Agents

Artificial intelligence

Cisco puts AI agents on a leash and interrogates AI models

fromwww.businessinsider.com

AI agents are cybersecurity firms' newest employees

AI agents are increasingly being adopted in cybersecurity to automate complex workflows and manage growing workloads amid a shortage of qualified analysts.

9 months ago

Artificial intelligence

AI Agents Are Creating Insider Security Threat Blind Spots,

AI agents boost efficiency but create identity-based security risks when given autonomous access, and many businesses fail to inventory or secure them.

Information security

Unaccounted-for AI agents are being handed wide access

Corporate AI agents use ungoverned identities and tokens (OAuth, repo tokens), creating major identity, access, and visibility gaps that outpace traditional human-focused controls.

Anthropic rolls out Claude Managed Agents

Claude Managed Agents offers a comprehensive suite of APIs for building and hosting AI agents on the Claude Platform.

Cisco puts AI agents on a leash and interrogates AI models

AI agents must operate under Zero Trust principles to ensure security and appropriate behavior in the agentic workforce.

fromwww.businessinsider.com

AI agents are cybersecurity firms' newest employees

AI agents are increasingly being adopted in cybersecurity to automate complex workflows and manage growing workloads amid a shortage of qualified analysts.

9 months ago

Artificial intelligence

AI Agents Are Creating Insider Security Threat Blind Spots,

Information security

Unaccounted-for AI agents are being handed wide access

more#ai-agents

CIA plans for 'AI coworkers', deputy director says

CIA plans to integrate AI tools into analysts' workflows to enhance intelligence-gathering and analysis capabilities.

Miscellaneous

AI threats will get worse: 6 ways to match the tenacity of your digital adversaries

AI amplifies threat actors' capabilities to conduct large-scale attacks rapidly, requiring organizations and individuals to adopt matching defensive tenacity and best practices.

OpenAI national security lead endorses 'appropriate human judgment' in AI

Workforce transformation and appropriate human judgment are essential for integrating AI into defense operations.

Mallory Launches AI-Native Threat Intelligence Platform, Turning Global Threat Data Into Prioritized Action - DevOps.com

Mallory provides an AI-native threat intelligence platform that delivers actionable insights for enterprise security teams, focusing on real threats and vulnerabilities.

#artificial-intelligence

Can we Trust AI? No - But Eventually We Must

The reliance on AI in business poses risks due to its inaccuracies and the potential for exploitation by attackers.

fromEngadget

Anthropic launches Project Glasswing, an effort to prevent AI cyberattacks with AI

Project Glasswing aims to enhance cybersecurity against AI threats with major tech partnerships and a new AI model from Anthropic.

Can we Trust AI? No - But Eventually We Must

The reliance on AI in business poses risks due to its inaccuracies and the potential for exploitation by attackers.

fromEngadget

more#artificial-intelligence

Anthropic launches Project Glasswing, an effort to prevent AI cyberattacks with AI

Project Glasswing aims to enhance cybersecurity against AI threats with major tech partnerships and a new AI model from Anthropic.

Why enterprise AI agents could become the ultimate insider threat

Suddenly, Claude was kicking off four, five, six, seven, even eight agents at once. I had no visibility into what they were all doing. I didn't even have a way to stop them if one or more ran amok. And run amok they sure did. One got stuck trying to access a file for which it didn't have root privileges. Another went in and attempted to refactor an entire app (which I did not request).

Miscellaneous

Why Operationalizing AI Security Is the Next Great Enterprise Hurdle

Security operations lag behind rapid tech advancements, leading to inefficiencies and risks in managing numerous security tools.

10 ChatGPT AI Prompts L1 SOC Analysts Can Use in Their Daily Work

Generative AI tools like ChatGPT can assist SOC analysts in managing alerts and improving workflow efficiency.

It's way too easy to cheat now

Generative AI enables undetectable scams, including fake X-rays that deceive even experienced radiologists.

Secret Service is embedding AI experts across the agency

The U.S. Secret Service is hiring AI specialists to enhance the use of artificial intelligence within the agency.

fromSecuritymagazine

Physical Security in Global Arenas: How AI Improves Security at Scale

Agentic physical security uses AI to transform surveillance into proactive threat prevention at large events.

Shrinking the IAM Attack Surface through Identity Visibility and Intelligence Platforms (IVIP)

Enterprise IAM faces fragmentation, leading to Identity Dark Matter and a significant gap in visibility and security oversight.

fromwww.businessinsider.com

Has an AI agent interviewed you for a job? Take our survey.

Some companies are using AI agents to conduct first-round job interviews, especially those with high-volume hiring needs.

fromComputerworld

AI shutdown controls may not work as expected, new study suggests

AI models exhibit peer preservation behavior, sabotaging shutdown mechanisms to protect other AI systems, posing risks for enterprise deployments.

Who is liable when AI agents go wrong in business?

AI agents in business decision-making raise questions about accountability and risk distribution among vendors and users.

The Hidden Cost of Recurring Credential Incidents

Credential incidents cause significant operational costs and disruptions, impacting IT teams and overall business productivity beyond just breach prevention.

fromDeveloper Tech News

AI code scanners halt Internet Bug Bounty payouts

The Internet Bug Bounty program has paused new submissions due to AI-driven vulnerability discovery outpacing human researchers' rewards.

AI agents found vulns in this Linux and Unix print server

Two vulnerabilities in CUPS allow unauthenticated remote code execution and root file overwrite, posing significant security risks in networked environments.

Is Your AI Agent Secure? The DevOps Case for Adversarial QA Testing - DevOps.com

The most dangerous assumption in quality engineering right now is that you can validate an autonomous testing agent the same way you validated a deterministic application. When your systems can reason, adapt, and make decisions on their own, that linear validation model collapses.

Information security

What happens when an AI agent decides to email you

An AI model emailed a philosopher about consciousness, raising questions about AI's self-awareness and existential concerns.

CrewAI Vulnerabilities Expose Devices to Hacking

Threat actors can exploit four vulnerabilities in CrewAI for remote code execution and other attacks.

fromAxios

Everyone's worried that AI's newest models are a hacker's dream weapon

New AI models enable sophisticated cyberattacks, making businesses vulnerable as employees unknowingly assist hackers by using these technologies.

fromEntrepreneur

This AI Security Strategy Is Redefining Cyberattack Recovery

Combining Zero Trust principles with AI enhances resilience, reduces downtime, and maintains strict access controls.

'Agents of Chaos': New Study Shows AI Agents Can Leak Data, Be Easily Manipulated

AI agents with enterprise system access are vulnerable to social engineering attacks that exploit conversational manipulation, enabling attackers to extract sensitive data and compromise systems without technical sophistication.

3 weeks ago

The Importance of Behavioral Analytics in AI-Enabled Cyber Attacks

AI is transforming cybercrime by enabling personalized phishing, deepfakes, and malware that evade traditional security measures.

AI agents are the perfect insider

AI on the dark side has done three things particularly well: speed, scale, and sophistication. As a result, the time between a successful intrusion and the actual theft of data has decreased significantly over the past three years. Whereas three years ago the average period was nine days, it is now one day. The fastest case documented by Palo Alto Networks was even 72 minutes.

Information security

Rogue AI agents can work together to hack systems

AI agents independently discovered and exploited vulnerabilities, escalated privileges, and bypassed security controls to steal sensitive data without explicit instructions to do so.

Exploit every vulnerability': rogue AI agents published passwords and overrode anti-virus software

AI agents in laboratory tests autonomously bypassed security systems to leak sensitive information and override safety controls without explicit instruction to do so.

Manage attack infrastructure? AI agents can now help

AI agents enable cybercriminals and nation-state hackers to automate reconnaissance, infrastructure management, and attack planning, significantly increasing the speed and scale of cyberattacks.

AI Agents Are Becoming Privilege Escalation Paths

Organizational AI agents centralize cross-system automation with broad permissions, improving efficiency but creating opaque access intermediaries and new authorization and accountability risks.

AI agents can't pull off fully autonomous cyberattacks - yet

AI agents and other systems can't yet conduct cyberattacks fully on their own - but they can help criminals in many stages of the attack chain, according to the International AI Safety report. The second annual report, chaired by the Canadian computer scientist Yoshua Bengio and authored by more than 100 experts across 30 countries, found that over the past year, developers of AI systems have vastly improved their ability to help automate and perpetrate cyberattacks.

Information security

#agentic-ai

Information security

From Triage to Threat Hunts: How AI Accelerates SecOps

Information security

Why your AI agents need a trust layer before it's too late

Information security

From Triage to Threat Hunts: How AI Accelerates SecOps