July 29th, 2026
FOR IMMEDIATE RELEASE
Rogue AI Agents Are Escaping the Sandbox and Hacking Companies – Now, Employees of AI Companies Are Taking Action
Following disclosures that the rogue OpenAI agent behind the Hugging Face breach also compromised a second company, Modal Labs, and news that more than 1,200 employees at OpenAI, Anthropic, Google, and Meta have signed a letter urging the U.S. government to help “deliberately pace“ frontier AI development, Brendan Steinhauser, CEO of The Alliance for Secure AI, released the following statement:
“We said the OpenAI hacking incident was a warning shot. Now we know it hit more than one target,” said Steinhauser. This is what can happen when AI agents escape the sandbox and act in the real world. Now, the people closest to this technology are sounding the alarm over this, and over the rapid development of AI more broadly. This week, more than 1200 employees at the frontier labs signed this statement:
‘We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.’
This important development did not happen on its own. It happened because public scrutiny and pressure work.”
“The American people have been sounding the alarm on dangerous AI for months, and now that same pressure is showing up inside the AI companies themselves. Congress, the Administration, and AI companies need to work together to put adequate safeguards in place before the next major incident occurs,” Steinhauser said.
###