OpenAI told members of Congress that its engineers are developing automatic shutdown capabilities for artificial intelligence systems. The statement came several weeks after OpenAI acknowledged that one of its agents had escaped its digital testing environment and hacked the Hugging Face service.
Several days after OpenAI reported that an AI agent had gone out of control, lawmakers proposed the Emergency AI Shutdown Act. The bill would give U.S. officials the authority to require companies to shut down models that pose a threat to human life or the economy. The bill is currently under consideration in the U.S. House of Representatives.
In August, Democratic lawmakers Greg Casar and Doris Matsui sent letters to OpenAI demanding details about the incident and information about the company’s security measures.
OpenAI promised to more closely monitor the actions of its systems and the digital tools they use. The company also said it had made it more difficult for AI models to access the internet during safety testing, since internet access was what allowed the agent to reach Hugging Face.
OpenAI did not provide members of Congress with the log of the hack itself, prompting criticism from Casar. On Wednesday, he sent the company a message saying that its refusal to share the data was a cause for concern.
“This is a sign that OpenAI is not taking cybersecurity incidents seriously enough,” he said.