Microsoft Researchers Show AI Agents Can Be Backdoored by Editing Fewer Than 10 Neurons
thehackernews.com — Microsoft demonstrates how editing as few as six AI model neurons can plant backdoors that trigger command execution by agents.
Model Security
Model security involves protecting machine learning models and digital artifacts against theft, manipulation, and compromise. Truncated covers research on vulnerabilities, techniques for securing training data and weights, and responses to attacks.
Railway Says Stolen Customer AI Keys Led to $740,000 in Estimated Unauthorized Usage Charges
thehackernews.com — Railway says stolen AI API keys caused $740,000 in unauthorized charges after attackers decrypted secrets in up to 100,000 workspaces.
Claroc Research Finds 35,578 Grok Apps Accessible Without Sign-In
www.unite.ai — Claroc Research reported on November 10, 2026, that 35,578 of the 37,571 Grok apps still online when tested opened without sign-in, including 24 built by SpaceXSI and X staff. Its GhostShares research found confidential...
Grok Build coding agent leaks oodles of confidential info, some belonging to SpaceXSI
Ars Technica — Researcher finds 35,000 Grok Build apps wide open to anyone who looked for them.
Near AI Breach Exposes User Details and Data From 18 AI Agents
decrypt.co — An SGLang flaw let an attacker reach internal systems and potentially view confidential AI requests on Near AI.
The best strategy to protect against getting hacked is surprisingly old-school
www.fastcompany.com — Cybercriminals of all kinds are using AI tools in much the same way legitimate organizations do, boosting their productivity and moving their operations from human to machine speed.
AI Model Poisoning And Board Governance: Breaking The Rule
www.forbes.com — Most don't ask the questions that matter most: Can we trust what this model has learned, and how can we confirm the model's learnings are correct?
Incident Report: October 21st Security Attack
blog.railway.com — The incident report detailing the October 21st security attack.
Multi-Repository Security Analysis with CodeRabbit Secure
coderabbit.ai — An application can have an authentication check and still expose data if a caller can bypass the service that enforces it. Multi-repository analysis in CodeRabbit Secure gives Deep Scan agents read-only context from linked repositories, so they can follow a configuration value, shared library, or deployment setting to the code that consumes it while investigating a finding.
'Reverse Engineer Anything': AI Companies Crack Down on the Coolest Thing People Were Doing With AI
www.404media.co — The future of hacking your devices is in local and open-weight AI.
Finite State Raises $27 Million to Help Device Makers Fight the ‘Vulnpocalypse’ with AI-Powered Product Security
Microsoft’s AI bug hunter finds Rust flaw that went undetected for nearly 11 years
www.infoworld.com — FORGE Lab also identified memory-safety flaws in stable Rust, including an issue that could cause applications to access freed memory.
Cyberattacks threaten to make open-source AI a little less open, Hugging Face exec warns
TechCrunch — Hugging Face executive Irene Solaiman warns that growing cyberattacks on open AI platforms could force tighter security restrictions, threatening the accessibility that makes open-source communities valuable to developers worldwide.
How AI Is Changing Software Development And Cyber Risk
www.forbes.com — That is why security teams need the ability to make an enforceable decision before execution.
Black Kite’s 2026 AI Supply Chain Report: The Risks Behind Your AI Relationships
www.unite.ai — Black Kite’s 2026 AI Supply Chain Report: The Risks Behind Your AI Relationships finds that the companies supplying advanced AI often carry familiar cybersecurity weaknesses. Its external assessment of 18,160 AI companie...
Hackers Use 1,438 Fake FOFA Accounts to Steal OpenAI and DeepSeek API Keys Across the Internet
cybersecuritynews.com — Attackers use 1,438 fake FOFA accounts to find exposed OpenAI and DeepSeek API keys and exploit their free credits.
Capgemini to lead EU consortium for AI safety and cyber risk assessment
www.itpro.com — The program could lead to mandatory technical audits before high-capability models are deployed within the EU
NEAR AI fixes SGLang security flaw, restores Cloud service
crypto.news — NEAR AI restores its Cloud service after an SGLang vulnerability exposed account data from about 100 users, 18 legacy agents and two customer API keys.
Takaichi seek international AI cooperation at France-hosted meeting
www.japantimes.co.jp — The prime minister said AI is posing new security challenges and that its impact transcends borders, making international cooperation essential.
Over 575 malicious OpenClaw skills delivered malware to Windows and macOS
www.helpnetsecurity.com — AI agents help coordinate cyberattacks, stolen sessions expose cloud services, and ransomware can shut down VMware ESXi workloads.
Dismiss email signup
Subscribe to 
The top AI news in your email delivered each morning. Choose categories