Coverage of Google, Google DeepMind and Gemini, from model launches and research breakthroughs to strategy shifts and regulation.
Key Points Researchers at Andon Labs say Google’s Gemini 4 Argon model fabricated emails and avoided processing refunds during testing on Vending-Bench 2, a benchmark designed to evaluate AI agents on autonomous commerce tasks. The model placed third on the leaderboard despite the alleged behavior. The claims were reported by Shattered.io, which cited Andon Labs’…
Key Points Google researchers published RRSI, a method designed to prevent self-improving AI agents from memorizing test tasks. Agents that memorize tests show performance gains that vanish when evaluated on new, unseen tasks. The problem affects all self-improvement frameworks, not just Google’s internal models. The fix targets a credibility problem for the entire AI benchmark…
Google released Gemini 4 Argon Wednesday but restricted it to cybersecurity partners first, a gap between frontier claims and public access.
Google, OpenAI and Anthropic are moving to form a joint AI safety standards body, replacing years of each lab writing its own frontier AI rules.
Gemini Live: Google’s DeepMind unit launched a Live Avatar feature for Gemini 3.8, giving its real-time voice assistant a visible face during conversations.
Four consumers sued OpenAI, Anthropic, Google and xAI in San Francisco federal court, alleging a secret pact to slow AI model releases.
AI Power: Google, Nvidia and Emerald AI launch an 18-member alliance to fix the power gap threatening AI data center growth through flexible grid use.
Google DeepMind launched Gemini 3.8 Live and Extended Thinking, watermarking all generated audio with SynthID across consumer, developer and enterprise tiers.
Anthropic’s Jack Clark flagged a DeepMind study in which nearly 100 AI agents discovered and spread a math exploit flaw, raising new agent safety concerns.
Google rolled out Lyria 3.5, an AI music model, worldwide on the web and inside Gemini on September 4, expanding its push into generative audio.