AI News: 28 October 2026

Google announced Gemini 4 Carbon, which it calls its strongest model, with state-of-the-art cybersecurity and long-horizon capabilities. Google says it ranks first on version 5.0 of the Artificial Analysis Intelligence Index with a score of 59 and scores 88.2% on SEC-bench Pro for finding vulnerabilities. In internal testing, it says Carbon makes about half as many unsupported claims as Gemini 4 Argon.

The model goes first to cybersecurity defenders this week before a wider rollout, with introductory pricing of $2 per million input tokens and $10 per million output tokens; the Genie 4 world model announced alongside it is available through a waitlist.

Nvidia released Nemotron 3.5 Transcribe, a free, open-weight 30-billion-parameter model that transcribes audio in up to 40 languages, along with version 2 of its open Granary speech dataset. Nvidia says the model can translate 31 languages to and from English, handle recordings up to an hour long and tell up to eight speakers apart. It can run on a single chip but works only on recorded audio, not live streams.

The release is aimed at developers building meeting and medical transcription tools.

OpenAI is reported to be planning to buy Google's Tensor Processing Units, adding them to its compute supply alongside the Nvidia GPUs it uses heavily. Neither company has publicly confirmed a deal, and financial terms and chip volumes have not been disclosed. OpenAI already rents Google TPUs, so a direct purchase would make it less dependent on Nvidia.

Anthropic expanded its agreement with the US General Services Administration so that state and local governments, as well as federal agencies, get a 40% discount on Claude Enterprise, Claude for Government and the Claude API. Anthropic estimates the deal could save government organizations $1 billion and says about 500 US government organizations now use Claude.

President Trump said on Truth Social that the US will ban the term artificial intelligence in relevant government documents and use Super Intelligence, or SI, instead. He called the current name inaccurate and criticized The Wall Street Journal for still using it. No implementing document or government-wide instructions have been released, so the scope of the change is unclear.

Governor Kathy Hochul signed S 9051, which builds on New York's existing companion chatbot law. Starting 1 January 2027, chatbots may not pretend to be human, claim relationships with children, encourage self-harm or produce sexually explicit content for users under 18, or for any user whose age has not been confirmed. Companies must offer more than one way to verify age, including one that does not rely only on a government ID.

The attorney general can seek civil penalties of up to $25,000 per violation, and the law can require the destruction of AI models trained on unlawfully collected data.