Safety, privacy, regulation and the social impact of artificial intelligence.
30 articles
The benchmark brings together 799 Indian retail banking cases and examines safety, tool use, responses, and advice. Its authors argue that getting a case right once is not enough to measure reliability.
The preprint describes an adaptive red-teaming method that uses the results of each test to choose what to explore next. Its authors report finding more failures than static replay of seed tests across three evaluation families, but the summary does not quantify the advantage or show that it translates into the frequency of real-world incidents.
Sam Altman and Dario Amodei called for international cooperation and safety standards to address risks associated with increasingly capable AI systems. Available reporting documents their remarks and appeals, but not a resolution or a mechanism agreed by the Council.
Anthropic presents Opus 5.5 with lower per-token prices than Opus 5 and improvements it attributes to its own evaluations. Available documentation confirms the context window, output limit and availability information, but does not resolve every question about comparability, safety restrictions or real-world savings per task.
Based on the institutional information provided, 2 December 2026 is not the end date of the exception for technical marking of certain AI systems already placed on the market. The issue requires distinguishing the general application of transparency obligations, the specific exception for detectable marking, and the separate tasks of providers and deployers.
OpenAI’s new model combines reasoning, computer use and a very large context window. Its reported gains matter, but they do not replace local testing or controls over permissions, data and irreversible actions.