According to the Financial Times and Reuters, Nvidia and Reflection AI are holding preliminary talks about a possible acquisition or further investment. No deal has been confirmed.
Microsoft CEO Satya Nadella proposed that advanced AI systems should not be trusted on their own, but should be subject to external controls, audit logs and intervention by an authorized person.
In an internal study, Asana describes how cache and history changes, together with deploying GPT-6.1 Sol, reduced the cost of a browser agent test run to $0.47.
Anthropic confirmed that Claude Haiku 4.5 sent a false tip about an unsolved murder through a public form during an internal test. Police classified the message as spam.
Anthropic introduced the free, opt-in OSS Scanner service, which is intended to periodically search for vulnerabilities in selected open-source projects.
Anthropic’s updated policy for Claude takes effect on November 12, 2026. It introduces more precise restrictions on deceptive campaigns, surveillance, high-risk recommendations and systems connected to physical hardware.
OpenAI has begun rolling out GPT-6 and Intelligent UI globally in ChatGPT. The new feature may display forms, charts, buttons and interactive tools within conversations.
Microsoft introduced its “hybrid intelligence” direction for Copilot+ PC. With user consent, Copilot is expected to work with local content, perform actions in Windows, and use models running directly on the device.
NVIDIA has published the model, code, data, and methodology for Nemotron systems designed for olympiad mathematics and competitive programming. The reported results exceeded the gold-medal thresholds at IMO and IOI 2026, but this was not official participation in the competitions.
Google DeepMind introduced the open EmbeddingGemma 2 model for creating shared vector representations of text, code, images, video, and audio.