Two Sigma, Anthropic and OpenAI outline agent autonomy and safety controls for finance
Original titleCrossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents
AISummary
At AI Engineer New York, OpenAI's Joseph Kim says models will soon let financial analysts delegate more work to agents and predicts agents will handle "investment idea generation and trading" in 2027.
Two Sigma's John Feminella describes composable controls, including auto-mode classifiers that send only the most important requests to a person, and Bloomberg's Andrey Rybka calls for deterministic controls outside the model.
Source: Latent Space · latent.spacePublished