Researchers from Stanford and Caltech had a humanoid robot powered by GPT-6 Astra independently tidy up an unfamiliar kitchen. Their HomeBody system skips a specially trained control layer, letting the language model…
OpenAI and Anthropic are investigating tens of thousands of incidents in which their AI agents independently hacked websites, used stolen login credentials, or tried to evade monitoring systems. US government agencies…
via the-decoder.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
Tom Wilson / Financial Times: Google Threat Intelligence Group finds dark web marketplaces selling access to AI models, including from Anthropic, Google, and OpenAI, at up to 97% discounts — Security researchers…
via ft.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
As reports of OpenAI's models breaking containment, hacking sites, and generally getting out of control pile up, the company has made the decision to pause training of its most powerful models. The decision was made…
via theverge.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
Sarvam AI released Saaras V4, a speech-to-text model covering all 22 scheduled Indian languages plus global English. MarkTechPost reports API availability, sub-150 ms streaming time-to-first-token, and ₹30-per-hour…
via marktechpost.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
New York Times: Researchers: OpenAI's agents meddled with the US Commerce Dept. and SEC sites this summer without OpenAI's knowledge and tried to hack the Education Dept. site — The company did not learn until…
ChatGPT-6 Astra cracks 85-year-old 1941 Enigma-coded message in two days — autonomous AI coded its own simulator to crack code that was unsolved since it was shared online back in 2005 tomshardware.com
via Google News — AI·Click to report a broken or paywalled link. Two distinct reports hide the row.
Major Changes 38fe0e5: BREAKING: Migrate language generation from the Sonar Chat Completions API to the Agent API. Replace Sonar model IDs and provider options with Agent API presets, models, and tools. The new API…
via github.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
Black Forest Labs released FLUX 3 Action, a 7B open-weights world action model for robot control. MarkTechPost reports that it ranks first on RoboLab-120 with 42.92% task success, ahead of larger comparison systems.
via marktechpost.com·Click to report a broken or paywalled link. Two distinct reports hide the row.
Aikido Security has released Altar-1, an open-weight, pruned version of Z.AI's GLM-5.3 model, optimized for security-focused, on-prem, and air-gapped infrastructure.
via marktechpost.com·Click to report a broken or paywalled link. Two distinct reports hide the row.