AI Daily — July 26, 2026

Models & Research

Google ships Gemini 3.6 Flash and 3.5 Flash-Lite, and restricts a new cyber model to governments and trusted partners — Google announced three Gemini models on July 21, 2026. 3.6 Flash is the new workhorse tier, priced below 3.5 Flash at $1.50 per million input tokens and $7.50 per million output tokens, with Artificial Analysis measuring 17% fewer output tokens than its predecessor. 3.5 Flash-Lite targets high throughput at $0.30 and $2.50 per million tokens and 350 output tokens per second. Both add computer use as a built-in tool. The third model, 3.5 Flash Cyber, patches vulnerabilities inside Google's CodeMender agent and will go only to governments and trusted partners in a limited pilot, which Google attributes to dual-use risk. Google also says Gemini 3.5 Pro is in partner testing and Gemini 4 pre-training has begun. Google ↗

My takeaway: Google has released three models. I was expecting the new flagship pro model, but these were budget models. However, based on the Artificial Analysis data, it is very interesting to find that the 3.6 Flash model is matching a prior Pro preview model. This also gives me an expectation of what the next flagship can do. If latency and cost are important factors for your service, I think these new Flash models can be good candidate to be evaluated.

Policy & Society

Thirty five AI companies sign a joint statement urging US policy support for open weight models — Microsoft published a joint statement on open weight AI on July 24, 2026, signed by 35 companies and organizations including OpenAI, Meta, NVIDIA, IBM, Hugging Face, Mistral, Mozilla and Y Combinator. The signatories argue that US AI leadership depends on a broad open ecosystem rather than any single frontier model, and that open weights widen access, increase competition, reduce provider lock in, and may improve security by spreading scrutiny across many teams. They acknowledge that released weights escape developer control but say prohibition is the wrong response. Their policy asks are more compute access for startups and researchers, public investment in shared datasets and evaluation tools, no premature restrictions on open models, and no treatment of distillation as misappropriation. Microsoft ↗

My takeaway: In my opinion, both open weight and closed weight models are needed, and competition between the two approaches can produce positive synergy in building better models. However, proper regulation and policy need to be set around data provenance and privacy. On distillation, the statement itself separates the technique from unlawful extraction of closed models and asks that extraction be handled through targeted legal frameworks. I agree with that split. Distillation that depends on unlawful extraction should not be defended as legitimate.

Summaries are AI-generated and may contain errors — always verify against the linked original. Each story links to its source, which holds the copyright. Outlet names are shown for attribution only and do not imply any endorsement or affiliation.

Disclaimer: The views expressed in My Takeaway are my own personal opinions and general observations on industry trends. They are not intended to criticize, disparage, or make factual claims about any specific company, product, or platform. Any platform names mentioned are referenced solely for illustrative and informational purposes.