Rosetta Intel

Product and model movements at the big three (OpenAI / Anthropic / Google DeepMind); 🛡 marks security-relevant items.

OpenAI

IPO in preparation— OpenAI is reportedly preparing a confidential listing filing, as early as September 2026, at a private valuation of about $730 billion.
In-house "Jalapeño" inference chipannounced; GPT-5.6expected to launch soon (oriented toward speed/reliability/accuracy).
ChatGPT Lockdown Mode(2026-02) remains the reference mitigation for AI browser/agent injection — OpenAI has acknowledged such injection "may never be fully fixed."

Claude Science released— Dario Amodei sets a goal of using AI to compress life-science R&D cycles by 10×, building on the acquisition of Coefficient Bioand the arrival of John Jumper(AlphaFold lead).
Claude Sonnet 5continues to roll out — quality approaching Opus 4.8, introductory pricing ($2/$10 per million tokens through 8/31), able to autonomously invoke the browser/terminal.
As agentic tool-invocation capabilities grow, the prompt-injection and abuse surface grows with them — boundaries and monitoring need to keep pace.

Gemini 3.5 Pro delayed to July 17— abandoning the original 2.5 Pro architecture for a complete rebuild, focused on mathematical reasoning, SVG scene generation, and image quality, positioned against GPT-5.6 and Anthropic Fable 5.
NanoBanana 2 Liteimage model released — generation in <4 seconds, from $0.034 per 1,000 images, with quality above the original NanoBanana.