DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
By Ryan Whitwam · Nov 10, 2026, 9:00 AM CST
Google DeepMind has revealed a new generation of the AI-powered Gemini Robotics platform. In the 2.5 release, Google says its robotic AI is getting tantalizingly close to completing real-world tasks as well as humans do. In fact, some Gemini robots were able to get actual work done in Mercedes-Benz factories, and the team is preparing for a day when robots operate alongside humans with an open sou
Excerpt shown under fair-use limits. Full text remains with the original publisher.
People in this coverage
Explore their history and attributable record. Being mentioned does not imply endorsement.
Layer 1 · Claims & fact checks
AI analysisLayer 3 · Reporting analysis
AI analysisEmotive languageThe phrase uses a subjective, hype‑driven descriptor that frames the technology as nearly human‑level without providing measurable evidence.
Future framingThe statement projects an optimistic future scenario, positioning the announcement as a step toward broader societal integration, but offers no concrete timeline or plan.
Authority appealThe article repeatedly attributes claims to DeepMind, relying on the organization’s reputation rather than independent verification.
Specific demonstrationProviding a concrete example of a dexterous task grounds the article in observable performance, supporting the claim of improved capability.
Context
AI analysisMissing context
The article does not provide details about the specific tasks performed in the Mercedes‑Benz factories, the criteria used in the 200‑activity evaluation, or any quantitative performance metrics comparing Gemini Robotics 2.5 to prior versions or to human baselines. It also lacks information about the scope, methodology, and accessibility of the claimed open‑source safety benchmark “ASIMOV‑jailbreak.”
Important context
DeepMind’s Gemini Robotics platform combines a vision‑language model (Gemini Embodied Reasoning) that interprets instructions from live video with a vision‑language‑action component that generates robot motions. The demonstration of tying a bow highlights progress on manipulating deformable objects, a known challenge in robotics.
Opinion vs. reporting
AI analysisThe piece mixes factual reporting (e.g., release name, evaluation count, specific demonstration) with evaluative language such as "tantalizingly close" and forward‑looking statements about future robot‑human collaboration, which are more opinion‑like and not substantiated within the text.