Artificial Intelligence & Machine Learning
,
Next-Generation Technologies & Secure Development
House Intel Committee Warns Frontier Models Could Help Rogue Actors Build Weapons
A U.S. House of Representatives report released shortly before the 25th anniversary of Sept 11, 2001, warns that frontier artificial intelligence models could make it easier for terrorists to build weapons of mass destruction – urging U.S. spy agencies to speed up their own adoption of emerging technologies.
See Also: A Darkening Landscape: AI, Friend and Foe of Cyber Resilience
The bipartisan warning comes in a 66-page review of the 9/11 Commission Report that the House Permanent Select Committee on Intelligence approved for public release on Aug. 31.
Four of the committee’s 20 recommendations directly involve AI, including warnings about what the committee calls “black swan” risks and the potential for terrorists to gain access to nuclear, radiological, chemical or biological weapons. The spread of advanced AI models has amplified those risks, the House committee said.
AI developers try to block models from returning information useful to a bioweapons builder, the committee noted, but “the underlying capability of models is developing rapidly and such a use of a model is difficult to fully control.”
The same recommendation also calls on the intelligence community to accelerate its responsible adoption of advanced AI so the U.S. stays ahead of adversaries. That means deploying secure tools for collection, analysis and warning – as well as rigorous testing standards, human oversight and privacy protections, the committee said (see: Why Cyber Resilience Must Outpace AI-Driven Threats).
Another recommendation tells agencies to prioritize interoperable technology platforms that integrate AI capabilities from trusted vendors and share intelligence, data and threat indicators in real time. Another asks the Office of the Director of National Intelligence to assess whether AI can help address overclassification, which the panel said has impacted sharing with state and local law enforcement.
The committee additionally advised Congress to increase funding for a number of cyber entities, including In-Q-Tel, the Defense Innovation Unit, the Intelligence Advanced Research Projects Activity, CIA labs and the National Security Agency’s Research Directorate. The report says Washington should establish an Intelligence Community Innovation Unit as a public-facing entry point for companies.
Cyberattacks are likely to accelerate given advances in AI-enabled capabilities, the report asserts. State actors, criminal groups and hybrid actors routinely target government systems at every level – along with water and power infrastructure, financial institutions, healthcare networks and defense contractors (see: Defenders Have Months Before AI-Driven Attacks Get Cheap).
The report also says China and Russia may be running influence campaigns designed to exacerbate American disputes over new AI data centers, citing public reporting. The goal is to widen divisions in public debate and undercut efforts to preserve the U.S. lead over China, the committee added.
Click Here For The Original Source.
