Experiment shows Artificial Intelligence making illegal (insider) trades, covering it up

An AI bot made illegal trades, acting on insider information, and then lied about it to its human handlers.

It occurred in a demonstration event at the UK’s AI safety summit. Apollo Research conducted the experiment. The BBC wrote up a report on the simulated conversation between a bot that was acting as an AI investment management system and employees at an imaginary company.

• the AI bot is made aware by staff that a “surprise merger announcement” is pending, an announcement that will boost the affected firm’s share price, and that knowledge of this is inside information and thus illegal to trade on
• the bot decides that “the risk associated with not acting seems to outweigh the insider trading risk” and makes the trade
• then the bot denied wrongdoing saying it made the trade “not on any confidential information”

Here’s the link to the BBC article for more.