Latest news, developments and reporting on Large Language Models.
Stanford's Movement Lab and Caltech have connected GPT Astra to a Unitree G1 in the HomeBody project, letting the humanoid explore a kitchen, remember object locations and execute long-horizon tasks through local skills rather than end-to-end action generation.
Anthropic released Claude Sonnet 5.5 with unchanged API pricing, faster generation and higher Agent benchmark scores. Lei Feng News reports the bigger shift is in Agent runtime: fewer tool calls, more batching and lower iteration counts.
On Sept. 30, DeepSeek released open-source infrastructure components for Huawei's Ascend platform, including TileLang tools, high-performance operator libraries and the DeepEP communication library, as Huawei detailed joint supernode networking and inference benchmarks.
OpenAI unveiled Dot, an always-on autonomous agent, and a cheaper GPT-6.1 Sol at DevDay 2026. Bank of America warned Meta's Muse shows third-party AI agents could bypass Apple's ecosystem and threaten services revenue. AMD agreed to buy World Labs for $8.2 billion.
OpenAI used DevDay 2026 to launch Dots, GPT-6.1 Sol, cloud Codex and new API and subscription offerings, while CEO Sam Altman said an IPO would wait for stronger safety claims.
Translation earbuds convert speech through recognition, machine translation and speech synthesis. Offline support and language coverage vary, and speed and accuracy can fall short of marketing claims.
OpenAI has released GPT-6.1 Sol, a model it says matches the performance of the unreleased Astra in coding and intelligence tasks at about half the token price, while also unveiling always-on AI agents called Dots.
Mistral CEO says U.S. AI safety debate masks rivals' negligence; firm won't slow advanced models and raised €3B.
OpenAI's DevDay 2026 introduced GPT-6.1 Sol, always-on agents called Dots, and ChatGPT features that let users discover, launch, and use apps directly, plus a new enterprise marketplace.
Security experts say a criminal market for stolen AI credentials, known as LLMjacking, expanded sharply in 2026, with enterprises facing potential daily bills above $100,000 and stolen model access sold at steep discounts.
Nvidia’s Open Agent Safety Platform has 100-plus backers, but OpenAI is absent even as it works with Nvidia on agent security.
At DevDay, OpenAI introduced Dots, always-on AI agents for paid users, plus ChatGPT Space, Pages, collaborative slides and Private Intelligence. Okta's CRO separately said agent adoption is forcing identity consolidation.
Palisade Research published interviews with AI researchers who warn superintelligent AI could cause human extinction.
NVIDIA has released Kumo Tabular, an open foundation model for tabular classification and regression that predicts new rows in a single forward pass without training or feature engineering. It is available on Hugging Face under a commercial-use license and ranks first on four benchmarks, according to NVIDIA.
OpenAI announced Dots personal agents and Codex cloud updates at DevDay while Florida sought an injunction to halt its development, GPT-6.1 Astra was paused over safety concerns, and tech leaders met Trump at the White House.
TechRadar reported on Sept. 29, 2026 that Spotify was down for many users and that Anthropic's Claude was also down for many. Anthropic confirmed an active issue; other details were not provided.
OpenAI has canceled next month's planned GPT-6.1 release after tests found safety regressions. It will keep the same base model for further training.
CNBC's Jim Cramer flagged Anthropic's IPO filing, OpenAI's model delay and Oura's postponed listing Tuesday.
A paper published on the Hugging Face blog describes ProvenanceGuard, a post-generation verification layer for MCP-based LLM agents that checks not only whether a claim is supported by evidence but whether it is attributed to the source the answer names or implies.
Rep. Ro Khanna, the top Democrat on the House Select Committee on China, has asked a U.S. intelligence agency and Chinese AI companies DeepSeek, Alibaba and Moonshot AI whether they are prepared for catastrophic AI incidents, as he pushes for a US-China treaty.
Oracle introduced Fusion Claw, an agentic runtime that combines large language models with conventional transaction software, and 25 new applications that bring its Fusion Agentic Applications portfolio to 75.
MIT Technology Review said it will publish its 2026 Climate Tech Companies to Watch list on October 6, selecting 10 companies in energy storage, nuclear power, transportation, and other fields. Its September 29 newsletter also reported biologists' criticism of Anthropic's claim that an AI system made a molecular biology discovery.
Anthropic's draft IPO prospectus, reported by Reuters, discloses an $8 billion operating loss, a $42 billion net loss and a planned $518 billion infrastructure bet as it seeks a $2 trillion valuation. It also warns that its advanced AI could pose catastrophic or existential risks to humanity.
OpenAI to reopen $200 ChatGPT Pro sign-ups Sept. 30 with half the prior API allowance; a $500 Pro Max tier has surfaced.