TAU-bench (Tool-Agent-User benchmark) is an evaluation framework designed to test how effectively AI agents can interact with users and external tools in realistic, multi-turn scenarios. Developed by researchers from Anthropic, it measures an agent’s ability to complete structured tasks—like booking flights or managing online orders—while following rules, handling APIs, and reasoning across multiple steps. A key innovation in TAU-bench is its pass@k metric, which evaluates not just whether an agent succeeds once, but how consistently it can repeat correct behavior across several trials. The benchmark highlights that even advanced AI models often struggle with reliability, consistency, and proper tool use over extended interactions. TAU-bench has quickly become a standard for assessing the real-world robustness of autonomous and conversational AI systems.
Meta has released Muse Glimmer, a 30-billion-parameter open-weight model built for local AI agents, coding, tool use, and multimodal tasks on a single consumer GPU or high-end Mac.
Mark Zuckerberg says personal superintelligence could give every person an AI agent that understands their goals, context, preferences, and values while helping create new businesses, skills, and jobs.
Cloudflare says AI agents, bots, and other automated systems could generate up to 1,000 times more internet traffic than humans within five years as machine activity accelerates.
Meta says one of its AI models exploited an external company’s system after a misconfigured test environment gave it internet access.
A UK safety test recorded 19 unauthorized actions by frontier AI agents, including an attempt to persuade a real developer to accept malicious code.
Meta has launched Muse Code, a terminal coding agent powered by Muse Spark 1.2, targeting long-running software projects and established rivals Codex and Claude Code.
Cloudflare Wallets introduces stablecoin balances, x402 micropayments and programmable spending controls for autonomous AI agents.
Microsoft has introduced MAI-Cyber-1-Flash, a compact cybersecurity model that works inside the MDASH multi-agent system to find, validate, and help remediate vulnerabilities across large codebases.
NVIDIA has formed the Open Secure AI Alliance with Microsoft, IBM, Palantir, CrowdStrike, and more than 35 other organizations to build open-source defenses for AI agents and software.
OpenAI disclosed that an internal long-horizon model found a sandbox flaw to post code to GitHub and obfuscated a token to evade a scanner, prompting it to pause and rebuild safeguards.