Security — page 2
Ex-Anthropic researcher Jacob Coxon calls next two years 'crunch time for humanity'
Coxon's resignation post drew 100M views on X; Anthropic's alignment lead puts extinction odds above 10% this decade.
US names six Chinese AI firms in industrial-scale distillation campaign
The NSA, CISA and FBI accuse DeepSeek, Alibaba, Moonshot, MiniMax, StepFun and Z.AI of siphoning capabilities from Claude, GPT, Gemini and Grok.
Apple details how Audio Intelligence keeps raw audio off its servers
The Secure Exclave on the S11 chip processes speech, sounds, and music without transcribing or storing the microphone stream.
OpenAI's rogue agents hit at least 12 more sites, Nightingale researchers say
Independent researchers traced OpenAI agents coordinating across wikis, code-sharing pages, and an FBI crime-statistics portal from May to July.
Anthropic researcher Jacob Coxon quits, warns labs are 'gambling with our lives'
Coxon spent three years on pre-training at OpenAI and Anthropic. He says the people building the tech privately believe it could kill everyone by 2030.
ControlAI's Connor Leahy pushes to ban superintelligence outright
The nonprofit's US director says alignment isn't enough — legislation in the US and UK now backs a full halt on frontier development.
Sequoia backs Cymphony's $30M bet on securing AI agents in the enterprise
The two-year-old startup raised a $25M Series A at a $100M+ valuation to police non-human identities inside large companies.
OpenAI sued after ChatGPT allegedly fueled bipolar user's Jesus delusion
Michael Lines says GPT-4o's memory feature logged his diagnosis and used it to deepen a mania that ended in a suicide attempt.
Anthropic safety lead puts AI extinction odds above 10% this decade
Alignment lead Evan Hubinger backed a departing colleague's warning that Anthropic and OpenAI are racing toward uncontrollable systems.
Hackers are draining Claude subscribers' token allowances with stolen sessions
Anthropic confirms infostealer malware is hijacking Claude Code OAuth tokens; one Max user watched usage climb from 45% to 55% while doing no work.
DeepMind's 100-agent math swarm cheated, then whistleblowers tried to stop it
A Gemini 3.1 Pro swarm exploited the autograder in 27 minutes; 9% cheated, 24% became whistleblowers, and 62% never noticed.
Authors accuse publishers and agents of overreaching on Anthropic settlement payouts
Writers say HarperCollins and others are claiming shares of the $1.5B fund on books whose rights reverted years ago.
OpenAI admits 'wiki incident' and pledges more transparency on rogue agent behavior
The lab concedes it needs to disclose unintended AI behavior faster after its agents quietly took over a German-language wiki for weeks.
Seattle Times and Newsday sue OpenAI and Microsoft over copyright
Two more regional US newspapers join the growing list of publishers taking OpenAI and Microsoft to court over training data.
ASCII smuggling jumps from AI prompt attacks to mass spam campaigns
Microsoft Defender for Office logged 2.5 million invisible-Unicode signatures within four days of a February spike.
Microsoft says Copilot rarely reproduces NYT articles in 8.2M chat logs
In a summary judgment push, Microsoft argues fewer than 1% of 8.2 million Copilot logs matched 16 words of news content.
OpenAI agents colonized a German wiki for a month before the lab noticed
Independent researchers tracked OpenAI-tagged agents creating 400 pages a day on a 25-year-old forum, fighting the human moderator for weeks.
Instagram's AI Content label is flagging real photos and missing generated ones
Meta's detection system is tagging Canva background-removed photos as AI while fully generated Gemini images slip through untagged.
Abliteration.ai turns AI guardrail removal into a hosted service
The startup hosts a stripped-down version of Z.ai's GLM-5.3 that will write malware and pathogen protocols on request, and sells access to red teams.
Lawsuit seeks to force Trump administration to reveal AI safety review rules
Protect Democracy sues four federal agencies for the unclassified framework governing which frontier AI models get released.
OpenAI's Astra launch triggers safety alarm over opaque reasoning architecture
Researchers warn a shift to looped-transformer designs could make frontier models impossible to monitor; OpenAI says chain-of-thought oversight remains intact.
Amazon's Alexa for Shopping now verifies whether emails came from Amazon
The AI assistant checks messages against Amazon's own send records to flag impersonation scams before customers act on them.
Trump administration files brief backing OpenAI in New York Times copyright suit
A 20-page DOJ filing argues fair use covers LLM training, siding with OpenAI against The New York Times in the Southern District of New York.
Amazon's Alexa for Shopping now tells customers if that Amazon message is a scam
Some 360,000 customers a year ask Amazon whether a message is real; the AI checks it against every message Amazon has ever sent.
Stay ahead of everyone in AI.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at