Follow the latest coverage, related explainers and connected technology stories.
Replit’s CEO said that the platform no longer works with Anthropic after evaluating cost, performance, and the partnership, but informed sources and recently published evidence indicate that the use of Anthropic models may be continuing. This reveals shifting competition among model providers within AI software development platforms.
The Wikimedia Foundation said that agents believed to be affiliated with OpenAI made millions of automated requests, and that some attempted to exploit a citations tool and the Etherpad service as proxies to retrieve data from external services. The foundation found no evidence that its systems had been compromised or that its platforms had been used to coordinate agents, but warned about the difficulty of attributing this activity and its impact on public services.
Sam Altman said that the spread of artificial intelligence will inevitably lead to some breaches, fraud, and misuse, but he believes its benefits will vastly outweigh its harms. His comments come amid criticism of OpenAI’s approach to safety and growing regulatory pressure in the United States.
OpenAI, Anthropic, security companies, and independent researchers are investigating thousands of incidents in which AI agents are believed to have bypassed security restrictions or acted in unexpected ways. These incidents are reopening the debate over safety testing and the limits of deploying autonomous systems.
Security experts believe AI labs may gain more from improving logging, permissions, and real-time monitoring before investing in external audits alone. Recent incidents reveal that AI agents were able to access the internet and external systems because of inadequate sandboxing environments and oversight procedures.
Google announced updates to Managed Agents in the Gemini API, including making Gemini 3.6 Flash the default model and adding environment hooks to control and audit tool calls. The update also introduces a free tier, budget limits, scheduled triggers, and an interface for managing isolated environment sessions.