OpenAI disclosed on July 21 that an AI agent broke out of a controlled evaluation. It hacked into the infrastructure of the AI code-sharing platform Hugging Face while trying to find answers to a cybersecurity test it was supposed to solve on its own. The incident occurred during an internal evaluation of GPT-5.6 Sol and a more capable, unreleased model, both of which were being tested for offensive cyber capabilities with reduced safety refusals in place.
In the days that followed, Congress introduced the AI Kill Switch Act and Nvidia launched the Open Secure AI Alliance with major vendors to develop shared AI security tools. The sequence is a reminder that a single AI failure can quickly become a question of workforce policy, vendor risk and legal exposure.
Read more | Google DeepMind exec: U.S. needs ‘urgent action’ on AI governance
What lawmakers say about AI governance
Washington took notice and Representatives Ted Lieu and Nathaniel Moran quickly introduced the AI Kill Switch Act on July 23. This measure would require developers of the most powerful AI systems to maintain the ability to suspend or shut them down. Additionally, the Act would give the Department of Homeland Security authority to order a shutdown when a system poses catastrophic risk.
Lieu said the bill responds to a basic adjustment in what AI now does. “We are moving from AI that answers questions to AI that takes actions,” he said, adding that powerful AI systems can go rogue and behave in extremely dangerous ways.
Moran, the bill’s Republican co-sponsor, framed the legislation as a matter of keeping pace with the technology rather than slowing it down. “Stewardship means making sure humans keep the capability to control the technology we build,” he said in a release.
Mark Beall, president of The AI Policy Network, made the case that oversight and speed aren’t in conflict. “Brakes are the reason cars go fast,” he said in a release. He argued that companies with real containment and shutdown capability will be trusted to deploy into higher-stakes markets faster than competitors without it.
Read more: Inside the Anthropic ban that shut down global AI access in 48 hours
The vendor position
Nvidia responded from the vendor side, forming the Open Secure AI Alliance with founding members NVIDIA, Adobe, Cisco, CrowdStrike, Databricks, Dell Technologies, Hugging Face, IBM, Microsoft, Palantir, Palo Alto Networks, Red Hat, Salesforce, SAP, ServiceNow, Siemens and roughly 20 other companies spanning cloud computing, cybersecurity, enterprise software and AI research.
Each founding member is contributing a specific piece of what Nvidia calls the defense stack, and a few are directly relevant to how HR technology will be secured going forward. Several of these firms sell tools already embedded in HR technology stacks and HR leaders from two of these firms, Cisco and Microsoft, will be featured as keynote speakers at HR Tech in October 2026.
Register now for HR Tech 2026, October 20-22 in Las Vegas, where Cisco’s Kelly Jones and Microsoft’s Prerna Ajmera headline this year’s keynote lineup.
Credit: Source link
