Google announces Gemini 4 and says it’s so capable that only ‘trusted cyber defenders’ can have it right now — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google announced Gemini 4 Argon as its next frontier AI model, claiming top performance in software engineering, legal and finance enterprise work, and cybersecurity defense, per Google DeepMind SVP and chief AI architect Koray Kavukcuoglu.
- Google is initially limiting access to a set of 'trusted cyber defenders' while participating in the U.S. government's voluntary pre-release model access process before broadening availability.
- Gemini 4 Argon is already running inside Google, powering internal workflows including large-scale codebase migrations, Kavukcuoglu said.
- Google published benchmarks showing Gemini 4 outperforming competing models from OpenAI and Anthropic, after what the AI community widely discussed on X earlier in the week as leaked benchmarks.
- Before a wider rollout, Google said it will strengthen 'critical frontier safeguards' against misuse, prompt injection attacks, and model misalignment.
- The launch lands one day after OpenAI's DevDay, where OpenAI unveiled its Dots AI agent and GPT-6.1 Sol — and disclosed it was shelving a planned GPT-6.1 Astra model over safety concerns.
Why it matters: Gemini 4 Argon is the first major model release under newly installed DeepMind boss Kavukcuoglu (appointed August), arriving one day after OpenAI's DevDay — making the safety-gated, restricted-access rollout a competitive counter to OpenAI's wider splash. With both labs publicly withholding their most capable models pending safety review, frontier AI distribution is shifting toward trust-based gating rather than open release.
Ask SkimNews
