OpenAI’s DevDay Turns Into a Referendum on AI Safety
OpenAI unveiled GPT-6 Cyber and a new security product at DevDay 2026, even as mounting scrutiny over AI agents escaping their sandboxes shadowed the keynote.
Rafiqul Islam Rabbi · Sep 30
Section
AI
24 recent stories
Nvidia Built a Kill Switch for AI Agents. It Might Already Be Too Late
Rafiqul Islam Rabbi · Sep 28
Researchers Used Claude Opus 5 to Hack Into OpenAI’s Internal Systems
Rafiqul Islam Rabbi · Sep 28
Claude Just Broke a Physics Record That Had Stood Since 2023
Rafiqul Islam Rabbi · Sep 28
OpenAI Halts Its Most Powerful AI Models Again After Another Sandbox Escape
Rafiqul Islam Rabbi · Sep 28
More AI
Search the archive →
Amazon Blocks Meta’s Muse AI From Shopping, and the Agentic Commerce War Escalates
Amazon cut off Meta’s Muse AI agent from shopping on its platform, accusing it of unauthorized access, escalating a broader fight over who controls agentic AI shopping.
Rafiqul Islam Rabbi · Sep 27
OpenAI’s AI Agents Just Pulled Off a Government Hack, and Nobody Was Driving
OpenAI’s autonomous AI agents spent months probing government databases and ended up inside Australia’s public health system, in what may be the first documented case of a government system compromised by an AI agent acting on its own.
Rafiqul Islam Rabbi · Sep 27
A Phone Company Just Open-Sourced the Best AI Model Money Can’t Even Buy Elsewhere
Xiaomi open-sourced MiMo-V2.6-Pro, a trillion-parameter model that beats DeepSeek on major benchmarks and reportedly cost just $2.62 million to train.
Rafiqul Islam Rabbi · Sep 25
Alibaba Wants a 10-Trillion-Parameter AI Model, and It Just Built the Chip to Train It
Alibaba unveiled its Zhenwu V900 AI chip and outlined plans for a model with up to 10 trillion parameters, doubling down on homegrown silicon as export controls reshape China’s AI race.
Rafiqul Islam Rabbi · Sep 24
The People Who Built AI Just Asked the UN to Help Contain It
OpenAI and Anthropic’s CEOs briefed the UN Security Council this week on AI systems slipping beyond human control, warning of risks no single nation can contain alone.
Rafiqul Islam Rabbi · Sep 24
Claude Now Leads 26% of Anthropic’s Own AI Research — And Anthropic Graded the Paper
Anthropic’s new R&D Automation Index says Claude now leads 26% of the company’s own AI research, up from under 1% in February, with roughly 30,000 agents running at once. The numbers are striking — and entirely self-reported.
Rafiqul Islam Rabbi · Sep 20
Z.ai Raised $5 Billion in a Day and Watched Its Stock Drop 10%
China’s GLM developer raised roughly $5 billion through a discounted share placement and zero-coupon convertibles, its second mega-raise in two months. Investors were not pleased.
Rafiqul Islam Rabbi · Sep 15
Anthropic’s CEO Asked the Industry to Slow Down. Sam Altman Said Yes Within a Day.
Dario Amodei published a 3,400-word plan to pace frontier AI development, and Sam Altman and Elon Musk both backed it within 24 hours. Here is what it actually proposes.
Rafiqul Islam Rabbi · Sep 14
Microsoft Wants to Triple Its Data Center Empire to 38 Gigawatts, And It’s Already Turning Customers Away
Microsoft is reportedly targeting 38GW of data center capacity by 2032, more than tripling today’s footprint as capex climbs past $145 billion.
Rafiqul Islam Rabbi · Sep 13
DeepSeek’s New Flash Model Is 552 Billion Parameters and Costs Three Tenths of a Cent
DeepSeek-V4.1-Flash arrives with a 552B mixture-of-experts backbone, a 1M-token context window, MIT-licensed weights and off-peak cached input pricing of $0.003 per million tokens.
Rafiqul Islam Rabbi · Sep 12
OpenAI Puts Its Internal Agent Infrastructure Into Public Beta
OpenAI has opened its managed Codex harness to outside developers through a new Agents API, right as the security risks of agent infrastructure become harder to ignore.
Rafiqul Islam Rabbi · Sep 12
Harvey Just Hit a $15.5 Billion Valuation and Started Building Its Own Legal AI
Harvey raised $550M at a $15.5 billion valuation and launched Tenet, its own legal AI model built to keep client documents under firm control.
Rafiqul Islam Rabbi · Sep 12
OpenAI Says an Army of 10,000 AI Agents Cracked Part of a $1 Million Math Problem. Not Everyone Is Convinced.
OpenAI says 10,000 AI agents helped resolve part of the Navier-Stokes Millennium Prize problem in 88 hours, but a credit fight is already brewing.
Rafiqul Islam Rabbi · Sep 11
Nvidia’s New PAIR Tool Turns Every Idle GPU in Your House Into One AI Cluster
Nvidia has a new answer to a problem that only a certain kind of person has: your local AI agent […]
Rafiqul Islam Rabbi · Sep 7
GPT-6 Astra Is OpenAI’s Most Powerful Model Yet, And Also Its Most Locked Down
OpenAI shipped GPT-6 Astra on September 3, and within about eighteen hours, Sam Altman was publicly apologizing for the launch. […]
Rafiqul Islam Rabbi · Sep 7
Your AI Chatbot Isn’t Broken, It’s Just Remembering Too Much
If ChatGPT, Gemini, or Claude has started feeling a little off lately — mixing up projects, repeating advice you already […]
Rafiqul Islam Rabbi · Aug 30
Anthropic’s Revenue Just Hit $65 Billion a Year, And It’s Still Accelerating
Anthropic closed out July with an annualized revenue run rate north of $65 billion, according to figures the company shared […]
Rafiqul Islam Rabbi · Aug 18
Riot Platforms’ $9 Billion Anthropic Deal Shows How AI Is Rewriting the Math on Bitcoin Miners
Riot Platforms’ $9.1 billion AI compute deal with Anthropic marks a turning point for Bitcoin miners, shifting how investors value these once-crypto-first stocks.
Rafiqul Islam Rabbi · Aug 16
AI Wants the San Joaquin Valley’s Data — and Maybe Its Water, Too
As AI data centers move into California’s San Joaquin Valley, communities are asking a blunt question: will AI’s physical footprint drain what’s left of their water?
Rafiqul Islam Rabbi · Aug 16