Docs and sources: Google researchers worried about AI risks to children, including potential cognitive and emotional harm, as the company pushed AI into schools (Wall Street Journal)

Google is deploying its Gemini 4 Argon AI model to a select group of cybersecurity partners via Fairwind, under the U.S. government’s voluntary pre-release program. Employees report strong benchmark performance but limited real-world coding capability, a claim Google denies. Independent tests show Gemini 4 Argon matches GPT-6 Astra on the Artificial Analysis Intelligence Index and has a 15% hallucination rate versus Astra’s 51%.