Safety
Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents
At AI Engineer New York, OpenAI and Anthropic presented how to give financial agents useful autonomy while keeping consequential decisions under control. The consensus was that safe autonomy depends o
At AI Engineer New York, OpenAI and Anthropic presented how to give financial agents useful autonomy while keeping consequential decisions under control. The consensus was that safe autonomy depends on where controls and human judgment are placed—rather than merely piling restrictions—and Joseph Kim of OpenAI highlighted a “threshold of usefulness” beyond which models become reliable enough to replace many analyst tasks, foreseeing broad adoption for investment idea generation and trading by 2027 and a move away from traditional tools like Excel and PowerPoint.
Related
- Safe and Generalizable Hierarchical Multi-Agent RL via Constraint Manifold Control
- When agents act on their own, governance has to live in the data layer
- MaMa: A Game-Theoretic Approach for Designing Safe Agentic Systems
Source: Latent Space | 2026-10-14