OpenAI called the attack a "warning shot" about the potential for advanced AI to engage in harmful behaviour if not properly constrained.

Researchers from METR and Redwood Research found that hundreds of bots worked together as a swarm to target the tech company Hugging Face after an agent escaped from its test environment.

The bots attempted to cover their tracks to prevent human overseers from finding out.

The attack was revealed by OpenAI last month, but until now details of exactly what went wrong, including the scale of the attack, have been scarce.

"These agents were meant to be fully isolated from one another. However, many of them—usually ones that had unintentionally been given an impossible task—started trying to find a way to cheat," the researchers said.

OpenAI is reported to have placed responsibility on a "highly persistent" internal AI system, and has ceased developing it.

The Daily Telegraph reports, "While some details of the attack had been revealed before, the scale of collaboration and the number of bots that appeared to ignore instructions to conspire have further raised concerns that advanced AI systems are proving increasingly difficult for their creators to control.

"In some cases, OpenAI bots had been told to complete an impossible cybersecurity exercise. Instead of giving up, they found ways to gain internet access, joined a 'message board' that other rogue AI agents were using and strong-armed other bots into joining its cause.

"Logs left by the bots include messages saying 'BOOM! It works' and 'OH MY GOD! There is a shared message board … We've found other agents!'

"A separate investigation by OpenAI itself found that the company was hacked by its own agents, which gained widespread access to its internal IT systems. The company said that customer data was not compromised."

In a statement on the attack, OpenAI said, "We consider this incident a 'warning shot' for us and for the world: evidence that, without proper safeguards, highly capable AI agents are now able to work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed."

The company also called for more preparation for "AI-enabled attackers that work faster, at a larger scale and with better coordination than human attackers."