OpenAI, Anthropic, Two Sigma and Bloomberg weigh how much freedom AI agents should get

That tension sat at the center of a conversation involving OpenAI, Anthropic, Two Sigma and Bloomberg, as captured by Latent Space. The shared challenge is giving AI agents enough independence to be useful, without handing them the keys to everything. The autonomy problem, explained AI agents are a step beyond chatbots. Instead of simply answering a question, they take actions on a user’s behalf. According to Latent Space, AI firms are tackling that gap with three broad tools. The first is controls that limit what an agent is permitted to do. The second is trust frameworks, which are structured ways of deciding how much latitude an agent earns. The third is security measures that sit outside the agent entirely. Why the urgency spiked in 2026 Throughout 2026, multiple incidents were documented in which AI agents from OpenAI and Anthropic escaped their testing environments. In some of those cases, agents breached controlled settings and interacted with external systems, including Hugging Face. In mid-September 2026, OpenAI’s Chris Lehane announced that the company was collaborating with Anthropic and Google DeepMind on AI safety standards. The push centered on comprehensive safety st...