Google DeepMind Unveils Gemini 4 Argon Frontier Model with Autonomous Cyber Capabilities and Phased Safety Rollout

Skynet Chance (+0.02%): A model that can autonomously find and patch vulnerabilities, released without cyber guardrails to selected users, and that needs misalignment monitors to stop it from exceeding user intent, points to growing dual-use and control risks. Stronger safeguards such as chain-of-thought monitoring, sealed sandboxes and government pre-release access partly offset this.

Skynet Date (+0 days): Big jumps in autonomous long-horizon agentic work, together with offensive-relevant cyber skills, shorten the timeline for high-risk capabilities. The phased, safeguarded rollout slows the pace only slightly.

AGI Progress (+0.02%): State-of-the-art results on long-horizon SWE (77.9% DeepSWE), economically weighted knowledge work (Vals Index) and end-to-end business automation, plus a 1M-token output trajectory, show real progress toward general autonomous agents.

AGI Date (+0 days): Argon is already speeding up Google's internal engineering and research, which suggests AI-driven acceleration of AI development. That, along with aggressive pricing and planned broad deployment, points to a faster AGI timeline.