OpenAI and Anthropic discuss safety limits for financial agents
What happened
At the AIE New York conference this week, OpenAI and Anthropic laid out their frameworks for financial agent autonomy and safety, stressing human review, permission settings and trust boundaries. OpenAI's Joseph Kim predicted that by 2027 agents will be able to generate investment ideas and trade, a forecast of future capability rather than shipped progress. Anthropic used a trading anomaly investigation as an example, where an agent proposes order changes that a human then approves.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- Latent SpaceCrossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agents
OpenAI 与 Anthropic 在本周 AIE New York 会议上阐述金融智能体的自主性与安全框架，强调人工审核、权限和信任边界。OpenAI 的 Joseph Kim 预测，2027 年智能体将能承担投资创意生成和交易；Anthropic 则以交易异常调查为例，让智能体提出订单修改建议并交由人工批准。
Heat over time
Not enough continuous observations to draw a trend yet.