Google reports that agentic AI threat activity has passed a key threshold, with hackers using AI agents to turn software patches into exploits. Meanwhile, OpenAI launched an agent that operates hospital software, with CVS and Cigna signing on, and Alibaba unveiled Qwen4-Max.
Archive
All published AI & tech news briefs
Sam Altman said OpenAI will not go public in 2026 over safety concerns, while AMI Labs raised $1.03B for world models, though its first product timing is unclear. Meanwhile, OpenAI debuted ChatGPT Wallet for AI agent payments, and a report warns that one hijacked MCP server can turn trusted agent tools against users.
OpenAI has shut its standalone robotics team and moved researchers into core groups, betting that general-purpose models will drive embodied AI. Meanwhile, the FBI warned that agentic AI hacking frameworks such as ARTEX are going mainstream, and LeCun's latest work is reshaping the AI scaling debate.
Money keeps flowing into AI: Mirendil raised $1B for model development, LangChain hit a $3B valuation, and OpenAI unveiled ChatGPT Wallet for agent purchases. At the same time, nonprofits sued OpenAI, Anthropic, Meta and Google over AI risks, while Hinton warned humans could lose control of AI within years.
Nonprofits have sued five AI giants, including Anthropic and Google, seeking to freeze new model development over "catastrophic" risks. Elsewhere, Yann LeCun launched AMI Labs in Paris after leaving Meta, and LangChain raised $300M to expand its AI agent toolkit.
Fractile raised a $720M Series C to tackle the AI inference memory bottleneck, while IBM and Microsoft teamed up on an agentic AI platform for enterprises. Meanwhile, safety concerns dominated: Hassabis warned of an AI "oil spill," and Anthropic's IPO filing flagged a key alarm as AI speeds up its own research.
Jailbreaks now appear within hours of every frontier model launch, while new models arrived from Anthropic, OpenAI, Meta and Google. Investors backed enterprise AI agent security, with Bay raising a $37M Series A and WhiteMatter a $20M Series A.
OpenAI has locked in a funding round valuing it at $1.25 trillion, while Anthropic's IPO filing warns that AI could pose "existential risks to humanity." Meanwhile, Democratic 2028 hopefuls are meeting AI safety advocates in San Francisco, and NVIDIA's GR00T 2 robot foundation model has surfaced in leaked slides.
Reports on AI agents moving from tools to autonomous actors, including experiments hinting at risky machine "teamwork" in companies, come as Anthropic warns of catastrophic risks from rogue AI and the DOJ sets up a Super Intelligence Task Force. Elsewhere, UNIST's EchoTrace shows LLM recommenders amplify bias through feedback loops, and Microsoft will keep Copilot's Claude processing inside the EU Data Boundary.
OpenAI reportedly reached a $1.25 trillion valuation after an $11 billion round that could grow to $20 billion, while Altman announced a partnership with ADNOC and G42 in the UAE. Meanwhile, Thinking Machines unveiled Inkling 2 Small as it reportedly pursues $5 billion in funding, and Anthropic added personal finance tools to Claude ahead of its Nasdaq listing.
OpenAI is reportedly securing $11B at a $1.25T valuation, while Sam Altman says the era of recursive self-improvement has begun, drawing sharp criticism. Meanwhile, Yann LeCun's JEPA work questions whether ever-bigger models are the answer, and Demis Hassabis warns society is "not ready" for what's coming.
LeCun's JEPA paper reports a 109M-parameter model outperforming a 72B one at matching a vision model, fueling debate over whether the industry has taken a wrong turn. Meanwhile, Anthropic warns its AI models are improving nearly twice as fast with AI's help, and Tencent launches its open-source Hy4 LLM.