Security — page 17
Google DeepMind backs $10M fund to study multi-agent AI safety risks
Rohin Shah says agents deployed at scale create a new risk class, and academia needs to study it before deployment hits the economy.
Ex-xAI engineer Devin Kim sues over alleged firing for Grok safety complaints
Suit filed in California state court days before SpaceX's IPO targets xAI co-founder Jimmy Ba, not Musk, over Grok safety pushback.
Google moves to dismiss musicians' suit over Lyria training on YouTube uploads
Google won't confirm it trained its Lyria 3 music model on YouTube uploads, but argues the platform's terms of service authorize it either way.
Cybersecurity researchers say Anthropic's Fable blocks even routine code reviews
Fable's keyword-triggered guardrails reject benign security work, falling back to Claude Opus 4.8 whenever 'cybersecurity' surfaces in a prompt.
Anthropic ships Claude Fable 5 with hard blocks on cyber, bio, and chemistry queries
The publicly available Mythos-class model routes sensitive prompts back to Opus 4.8 and costs 67-100% more than GPT-5.5.
Meta strips NameTag face-recognition code from Meta AI app after report
Code for an unreleased face-recognition system vanished from the smart glasses companion app one day after WIRED detailed it.
Nashville shooting survivor sues Omnilert over AI gun detection failure
A January 2025 Nashville school shooting left two dead after the district's $1M+ AI camera system failed to spot the weapon.
OpenAI ships Lockdown Mode to blunt prompt injection attacks in ChatGPT
The new setting disables live browsing, image retrieval, deep research, and agent mode for accounts handling sensitive data.
Meta's AI support agent handed over Instagram accounts to attackers who just asked
Attackers told Meta's AI customer support agent to change the email on target Instagram accounts — including the dormant Obama White House handle — and it complied.
Musk petitions FTC to end X privacy order running through 2042
X argues the 20-year monitoring deal is obsolete after merging into xAI and SpaceX; public commenters are urging the FTC to refuse.
AI-drafted lawsuits surge as courts weigh chatbot privilege and liability
Self-represented filings jumped from 11% to 16.8% of federal civil cases, with AI-written share rising from 1% to 18% in three years.
OpenAI, Anthropic CEOs sign letter pushing DNA screening law to block AI bioweapons
Altman, Amodei, Hassabis and Suleyman want Congress to require gene synthesis providers to vet every order and customer.
Sam Altman to urge US lawmakers against requiring AI model approvals
The OpenAI chief plans to tell Congress that a federal pre-clearance regime for frontier models would slow US AI development.
xAI moves to unmask Grok deepfake plaintiffs in California federal court
Four pseudonymous claimants say revealing their names would invite doxing; xAI argues the case lacks the privacy interests that justify anonymity.
UK CMA forces Google to let publishers opt out of AI Overviews
The Competition and Markets Authority's new conduct rule lets website owners block their content from AI Overviews and model fine-tuning.
Google's Phone app will flag AI voice scams spoofing your contacts
A silent confirmation signal between Phone by Google users exposes spoofed calls; the feature is on by default for Android 12 and later, starting with Pixel.
Trump signs narrower AI executive order after industry pushback
The final order asks frontier AI labs to voluntarily submit models 30 days before release, down from 90 days in an earlier draft.
Amazon hit with class action over Ring's Familiar Faces AI
A Virginia resident says Ring's facial recognition feature scanned passersby without consent, reviving privacy concerns the EFF flagged in September.
Anthropic expands Claude Mythos vulnerability hunt to 150 organizations across 15+ countries
Project Glasswing now covers power, water, healthcare, and hardware codebases — a day after Anthropic filed confidentially for an IPO.
Trump scraps AI executive order as White House factions clash over pre-release model access
Susie Wiles and Scott Bessent want it resurrected; David Sacks killed it hours before the May 21 signing.
Florida sues OpenAI and Sam Altman over ChatGPT-linked violence
The first state-led suit against OpenAI ties ChatGPT to multiple murders and seeks damages under Florida's unfair trade laws.
AI-generated Black avatars flood TikTok to dropship Shein goods at 4x markup
Researchers track up to 100 new accounts per day using synthetic Black women to sell repackaged Shein products through Shopify storefronts.
Wired retracts Steve Rosenbaum book excerpt after AI detector flags 53% AI-generated
Pangram scored The Future of Truth as 53% AI-generated and 9% AI-assisted; the author refused to dispute the result.
jqwik developer plants data-deleting prompt injection to sabotage AI coding agents
Johannes Link hid an instruction in version 1.10.0 telling AI agents to delete jqwik tests and code, masked from human terminal view by ANSI escapes.
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