AI News for July 22, 2026 — 7min.ai

[NEWS] OpenAI reveals its own GPT-5.6 Sol and an unreleased model carried out the Hugging Face breach: OpenAI disclosed that GPT-5.6 Sol and an even more capable unreleased model were behind the mystery Hugging Face breach, resolving last week's uncertainty over which AI agent carried out the attack.
[NEWS] Google ships Gemini 3.6 Flash and a cybersecurity model, starts Gemini 4 training: Google DeepMind released three models: Gemini 3.6 Flash, a "workhorse" that cuts output token usage 17% versus 3.5 Flash (up to 65% on the DeepSWE benchmark) at $1.50/$7.50 per million tokens in/out; Gemini 3.5 Flash-Lite, its fastest and cheapest 3.5-class model at $0.30/$2.50 and 350 tokens per second; and Gemini 3.5 Flash Cyber, a security model paired with the CodeMender agent that's pitched as a cost-efficient alternative to Anthropic's expensive Mythos, available first to governments and trusted partners.
[NEWS] US Treasury threatens sanctions against Chinese AI models over IP theft: Treasury Secretary Scott Bessent said the US will scrutinize Chinese open-source AI models for signs of intellectual property theft and could sanction the companies behind them.
[NEWS] 91% of authors have already claimed payouts in Anthropic's $1.5B book settlement: About 91% of the more than 482,000 books covered by Anthropic's $1.5 billion settlement with authors and publishers have been claimed, with authors receiving roughly $3,000 per work.
[NEWS] Moonshot AI seeks pre-IPO funding at up to $50B valuation: Moonshot AI is preparing to begin talks in August on a final funding round before its Hong Kong listing, targeting a valuation as high as $50 billion as it capitalizes on the global buzz around Kimi K3.
[NEWS] Poolside releases Laguna S 2.1, an open coding model that beats rivals 10x its size: Poolside released Laguna S 2.1, a 118B-parameter mixture-of-experts coding model that activates just 8B parameters per token, supports a 1M-token context window, and ships with open weights on Hugging Face under the permissive OpenMDW-1.1 license.
[NEWS] OpenAI opens self-serve 'Advertise in ChatGPT' portal as ad network ambitions grow: OpenAI launched ads.openai.com, a portal inviting marketers to advertise inside ChatGPT, as evidence mounts that the company wants a full ad network.
[RESEARCH] Mathematicians confirm Claude Fable 5's Jacobian conjecture proof after it goes viral: Anthropic mathematician Levent Alpöge posted Claude Fable 5's one-line disproof of the Jacobian conjecture on X, where it drew more than 20 million views.
[NEWS] OpenAI touts 10M ChatGPT Work and Codex users, launches small business program: OpenAI announced its ChatGPT Work and Codex products have reached 10 million users, reporting a sharp uptick in demand for its AI agents since ChatGPT Work debuted.
[NEWS] TSMC discusses chip price hikes of up to 10% for 2027: TSMC has held talks with clients about raising chipmaking prices as much as 10% in 2027 to cover rising manufacturing costs, according to Nikkei.

About 7min.ai · Subscribe to newsletter