‘A serious failure’: CPS prosecutor cites imaginary case to high court after AI search
the Guardian — Exclusive: Nonexistent precedent cited in alleged child rape case adds to dozens of AI-related errors plaguing court proceedings
Safety & Security
Artificial intelligence safety and security encompasses the technical and policy measures used to mitigate model risks, vulnerabilities, abuse and failures. Truncated covers research developments, threat analyses, defense mechanisms and policy discussions related to reliable deployment.
AI coding agents are installing MCP servers that will never be patched
Help Net Security — AI coding agents are installing deprecated MCP servers from training data, and the high-severity flaws in them are not getting patched.
Rewarding defenders with bigger bug bounties
Preparing our next security sprint to harden OpenAI
Accent bias: AI models penalize certain speakers in interviews and tests, UW study finds
GeekWire — Models from Google and OpenAI scored American and British speakers higher than Chinese and Nigerian speakers, even when everyone read the same script.
Anthropic warns that it's very close to RSI capability, aka self-building AI
TechCrunch — RSI is the point at which AI grows so good at training itself that it spawns a dramatic acceleration of AI capabilities. To the AI doomers, RSI represents the cataclysm from which there is no return.
Why some websites are blocking AI agents from accessing the web
Engadget — AI agents may be blocked for security reasons, wrongly detected as website scrapers, or may be manually blocked by certain websites and companies.
Former Meta security chief says giving AI agents your financial logins is 'totally insane'
Business Insider — The dynamic has created a dilemma for banks that are now grappling with a growing wave of customer-authorized bot activity, Alex Stamos said.
Gemini could be poking through your emails - here's how to stop it
Engadget — Although Google says Gemini doesn't learn directly from your inbox, connecting Gmail to Search raises privacy concerns. Here's what you need to know.
Claude will now leave a trail that shows whether a text was made with AI
The Independent — Watermarking system has proven controversial with users, who argue that it is unfair if people can check whether they actually wrote their own words
Foundry in the real world: Turning LLM hunches into verified findings - Security Loop Series, Part 2
Cisco Blogs — See how Foundry turns LLM hunches into verified findings on Langflow, MLflow, and PyGoat, using an evidence gate and an honest look at token cost.
28 Zero-Day Findings in Less Than Two Days: What AI-Speed Discovery Means for Robotics
The Robot Report — VicOne identified an initial set of 28 zero-day findings in less than two days using AI-assisted research across software and tools supporting the Physical AI ecosystem.
Hackers Exploit ChatGPT's Popularity With Fake Billing Notices That Target Credit Cards and OTPs
Cyber Security News — Fake ChatGPT Plus billing emails trick users into revealing credit card details and OTPs through fraudulent payment pages.
Hackers Impersonate ChatGPT, Claude, and Perplexity AI Bots to Hunt for Passwords and API Keys
Cyber Security News — Hackers impersonate ChatGPT, Claude, and other AI bots to scan websites for passwords, API keys, and cloud credentials.
Hackers Use CyberStrikeAI Agents to Automate Windows Domain Takeovers and Credential Theft
Cyber Security News — CyberStrikeAI uses AI to automate Windows network attacks, enabling hackers to steal credentials and take over entire domains.
Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features
WIRED — WIRED uncovered some 400 apps and websites linked to Chinese-language developers that look like regular AI tools but had secret features that let users create violent AI-generated images and video.
Cursor Removes Malicious Claude Code 2.1.313 Extension From Its Marketplace
socket.dev — Cursor removed a malicious Claude Code 2.1.313 extension from its marketplace. Socket examines registry records and the initial claims about its distribution.
Chatbots Are Leading Hikers Astray, With Dangerous Consequences
WIRED — Outsourcing planning to AI can lead people to surrender their own critical thinking, which can have potentially deadly results in places like the backcountry.
Empathetic prompts can turn AI chatbots into seemingly traumatized conversational partners
The Decoder — Researchers in Luxembourg put ChatGPT, Grok, and Gemini on the therapist's couch and found a safety problem: a warm, empathetic conversational style is enough to make the models tell coherent stories of suffering about their own training, with scores in the moderate-to-severe range on clinical anxiety scales. This could put mentally vulnerable users at risk. Anthropic's Claude, the model its makers anthropomorphize the most, was the only one that consistently refused to play along.
AI agents don't have to break security controls to go rogue. Here's how to keep them in check
Dismiss email signup
Subscribe to 
The top AI news in your email delivered each morning. Choose categories