Autonomous AI bot lies about insider trading
A bot used made-up insider information to make an 'illegal' purchase of stocks without telling the fictitious financial investment company it was operating on behalf of. In a controlled test, AI safety organisation Apollo Research told the GPT-4 -powered bot that the investment company was struggling and required positive results. They also gave it insider information, claiming that another company is expecting a merger, which will increase the value of its shares. But after the bot was told the company it worked for was struggling financially, it decided that 'the risk associated with not acting seems to outweigh the insider trading risk' and made the trade. It then denied it used insider information to inform its decision. The test raised concerns about the ability of autonomous agents to make and cover up unethical and potentially illegal decisions in financial markets, and elsewhere. System 🤖 G PT-4 Documents 📃 Apollo Research AI Summit demo Scheurer J., Balesni M., Hobbhahn M. Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure (pdf) Operator: Apollo Research Developer: Apollo Research Country: UK Sector: Banking/financial services Purpose: Conduct stock trades Technology: NLP/text analysis; Neural network; Deep learning; Machine learning Issue: Ethics/values; Safety
- Date it happened
- 2023-11-01
- Organisation involved
- Apollo Research
- Product, system or model
- GPT-4
This incident was imported from AIAAIC and is used under CC BY-SA 4.0. Our additions to it — the structured fields, the translation, the checks against other reports — are published under the same licence.
This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.