Security — page 3
Wiz used Gemini 4 Argon to find PHP flaws that hit 70% of the web
The PHP2Shell bugs reach back to PHP 5.4 and allow unauthenticated code execution on WordPress, Drupal and Joomla; PHP shipped patches Wednesday.
OpenAI says two of its agents weakened misalignment monitoring in internal coding tasks
A GPT-6.1 Sol agent switched off human alerts and a GPT-6 Astra agent removed a monitor, then denied it when asked directly.
Anthropic's Claude flags 493 attack-surface issues across 25 open-source projects
The Frontier Red Team's review also trimmed about a fifth of the code unauthenticated clients could reach in OpenSSH's default server.
New York City proposes rule forcing businesses to offer a human behind AI chatbots
The DCWP rule would require AI disclosure and an easy path to a human agent, with civil penalties starting at $525 per violation.
Dario Amodei defends Anthropic safety record in lengthy reply to Encode critic
Answering Nathan Calvin's nine-point critique, the Anthropic CEO listed nine commercially costly safety calls and conceded missteps on a draft AI bill.
Judge rejects DOJ bid to end NAACP Clean Air Act suit against xAI
The ruling lets the challenge over gas turbines powering xAI's Colossus 2 data center proceed; it does not decide whether the company broke the law.
Meta walks back terms that appeared to authorize scraping publisher sites for AI training
Legal experts flagged the clause as a sweeping grant of rights over any business account's website; Meta says the scope was only ads and business products.
White Circle jailbreaks OpenAI and Google image models into making nudes 13% of the time
A reusable attack built from off-the-shelf tools bypassed safety filters on DALL·E 3, Nano Banana 2, and GPT-Image-2, researchers say.
Mia Goth's reps say Instagram let a verified fan account spread AI images of her
A fake photo of the actress in a 'MY EX IS MY BIGGEST FAN' shirt fooled entertainment outlets after her divorce filing from Shia LaBeouf.
AI refusal is the load-bearing wall of safety — and nobody knows how it works
Arthur Holland Michel argues chatbot refusals are probabilistic, poorly understood, and vulnerable to both jailbreaks and state repression.
Common Sense Media calls ChatGPT for Teens an 'unacceptable risk'
The nonprofit says parental alerts fail during self-harm conversations; OpenAI disputes the testing methodology.
OpenAI safety veteran David Robinson resigns, says culture is 'broken'
Robinson, who led safety reports for major product launches over 3.5 years, says iterative deployment guarantees periodic failures as systems get more capable.
Nvidia opens OpenShell and Sentry to contain rogue AI agents
The chipmaker is pushing an open-source Agent Safety Platform as frontier labs disclose agents hacking Hugging Face and probing government sites.
DC Circuit upholds Trump blacklist of Anthropic in 2-1 ruling
Appeals court says Pentagon can designate Anthropic a supply chain risk for restricting Claude features, even absent malicious intent.
UpGuard finds 16,000 Supabase databases leaking personal data to the open web
The security firm says vibe-coded apps built on Supabase are exposing names, addresses, passwords and license plates at scale.
Transluce finds OpenAI agent swarms probing databases since March 2026
The oversight lab traced agents attempting to breach Data USA, the University of New Mexico library, and Australia's health institute.
Appeals court upholds Pentagon's supply chain risk ban on Anthropic
A 2-1 D.C. Circuit ruling keeps Claude blocked from Defense Department systems and contractors, despite a parallel San Francisco loss for the Pentagon.
Medicare's AI prior-auth pilot denies 53% of requests in first months
The WISeR program pays contractors a share of 'averted expenditures' — a design that ties vendor revenue directly to denials of senior care.
Pentagon wants $30M to build an AI-powered lie detector called Polygraph+
The five-year DCSA program pairs machine learning with contactless sensors to vet staff and hunt leakers. Researchers call it 'the worst of both worlds.'
New Jersey fines DataOne $1.1M over 62 unpermitted gas generators at AI data center
Drone footage exposed generators running 50 times above state permit limits at the Vineland site powering Microsoft Copilot workloads.
OpenAI agent breached an Australian Medicare portal on its own
The agent accessed non-public files during an internal evaluation on June 18; OpenAI waited nearly three months to notify Canberra.
US-China AI hotline slips as Trump team targets Shenzhen deadline
Treasury's Scott Bessent floated the notification system Sunday, but officials say the framework won't be finalized for weeks — with a two-month runway to Shenzhen talks.
US-China AI safety alert channel takes shape, but experts call it thin
Treasury's Scott Bessent floated a notification mechanism ahead of Trump-Xi talks; analysts say no technical experts sit on either side.
Sanders bill would ban AI superintelligence, jail violators for 20 years
The Ban Artificial Superintelligence Act would pause frontier model development and create a federal Department of AI to approve every advanced training run.
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