Anthropic’s “Project Deal” has demonstrated that AI agents can autonomously negotiate and close real-world transactions, but the experiment also surfaced a quiet, troubling asymmetry: not all AI representations are created equal. In December 2025, Anthropic transformed its San Francisco office into a live classified marketplace, a Craigslist-style platform with a critical twist. Rather than negotiating […] The post Claude AI Agents Close 186 Deals in Anthropic’s Marketplace Experiment appeared first on Cyber Security News.