Lawmakers push AI security audits after White House flags model safety concerns

U.S. lawmakers are moving to tighten oversight of powerful artificial intelligence systems, with bipartisan legislation that would require independent security audits for advanced models. The push comes as the White House is said to be monitoring a reported OpenAI safety incident that has renewed debate over how much control companies and governments should have over fast-moving AI tools. According to reporting cited by Express Tribune, the proposal would introduce mandatory outside reviews for powerful models before they reach the public or are deployed at scale. The goal is to identify major security weaknesses, misuse risks and other safety issues that internal company testing may miss. The legislation is being framed as a way to add accountability to a sector that has grown faster than the rules around it. The policy idea lands at a moment when AI regulation in the United States is still fragmented. Federal agencies have issued guidance, the White House has backed voluntary commitments from major AI firms, and some states have started drafting their own rules. But there is still no single nationwide standard for testing the most capable systems. That gap has left lawmakers from both parties under pressure to show they can respond to public concerns about safety without stifling innovation. Independent audits could become one of the more practical guardrails in that effort. Supporters argue that outside experts can help assess whether a model is vulnerable to jailbreaks, data leakage, deceptive behavior or other failures that might be difficult to catch in-house. Critics of the industry have long said that companies developing frontier AI have too much incentive to move quickly and too little incentive to slow down for rigorous scrutiny. At the same time, the details matter. The feed material available does not make clear which models would fall under the proposed threshold, how often audits would be required or which agencies would enforce the rules. It also remains unclear whether the legislation includes penalties for non-compliance or a timeline for implementation. Those unanswered questions will determine whether the proposal becomes a meaningful safeguard or simply another broad statement of intent. The reference to a possible AI “kill switch” reflects the heightened political tone around the issue, but the substance of the debate appears to be centered on control, testing and accountability rather than dramatic shutdown powers. For lawmakers, the challenge is to create standards that are strong enough to reduce risk while still workable for companies building systems that are increasingly embedded in search, customer service, software development and government workflows. For businesses using AI, the impact could be significant if the bill advances. Mandatory audits would likely add cost and delay, especially for smaller developers trying to compete with larger players. On the other hand, clearer rules could also help establish trust with regulators, enterprise customers and the public, which has become a major issue as AI systems are deployed in sensitive settings. For now, the main takeaway is that the policy conversation is shifting from broad warnings about AI risk to more concrete oversight tools. The White House’s attention to the reported OpenAI incident and the lawmakers’ audit proposal both suggest that safety concerns are no longer being treated as theoretical. Whether Congress can turn that concern into enforceable law is the question that now matters most.

Source: Express Tribune Latest - https://tribune.com.pk/story/2619873/white-house-monitors-openais-rogue-ai-incident-lawmakers-propose-kill-switch