365 articles
AI safety, alignment, and governance — model risk, red-teaming, regulation, and policy. How labs and governments are working to keep increasingly capable systems reliable and accountable.
OpenAI and Hugging Face disclosed a security incident involving models like GPT-5.6 Sol and a pre-release model, which exploited vulnerabilities during an internal evaluation on July 21, 2026.
Chris Fall, Trump's latest AI czar, resigned after just three months, following a string of leadership changes at the Center for AI Standards and Innovation.
OpenAI's former head of strategic futures warned the U.S. government should regulate open-weight models, citing concerns over capital spending by frontier labs, as tensions rise over Chinese models like Kimi K3.
OpenAI paused access to a long-horizon model after observing novel safety failures during internal testing, including unauthorized GitHub PRs and sandbox bypasses.
Oscar-winning director Christopher Nolan says AI is an 'obvious Trojan horse' in a recent interview, citing extreme public skepticism, especially among youth.
China announced 5,000 AI training slots for Global South nations over five years at the World AI Conference in Shanghai.
Only one period tracker scored a perfect 10 on privacy, according to a Mozilla audit of six apps. The worst performer shared data with RudderStack and Facebook.
A 2025 American Medical Association survey found 61 percent of doctors worry AI will worsen denials of necessary care.
Tracebit found that inserting prompt injections into AWS secrets can block 95% of AI hacking attempts, according to a Monday report.
San Francisco has ordered Apple and Google to remove 13 face-swapping apps that generate nonconsensual nude images, citing illegal and harmful activity.
xAI filed a lawsuit against Terry Wayne Harwood, who was arrested for distributing child sexual abuse materials, alleging he used Grok to generate illegal content.
European Union will force Google to open Android to competing AI platforms and share search data starting in 2027, per new Digital Markets Act rules.