DeepMind Unveils Gemini Robotics 2.5 and Open ASIMOV‑Jailbreak Benchmark
Original headline: DeepMind reveals Gemini Robotics 2.5 and open safety benchmark "ASIMOV-jailbreak"
DeepMind announced its latest model, Gemini Robotics 2.5, alongside an open safety benchmark called ASIMOV‑jailbreak. The benchmark is designed to assess jailbreak vulnerabilities, and the new model achieves up to twice the success rate on certain tasks compared with prior versions.
Publishing an open jailbreak benchmark helps the industry gauge and improve AI safety against adversarial prompts.
Research angleHow does the availability of jailbreak benchmarks influence developers' security decision‑making for AI models? — Protection Motivation Theory
Turn this security issue into a research project
HJITM works with researchers from study design through surveys, analysis and submission to international journals in generative AI and technology management. Bring us your research question.
This item is an AI-generated summary of reporting by Ars Technica. Please check the original article for accuracy. Copyright in the original article belongs to Ars Technica.