US Congress Intelligence Committee Warns AI Could Be a New 9/11-Scale Threat

The US House Intelligence Committee approved by voice vote a report that, for the first time, places artificial intelligence alongside cyberattacks, drones and biological weapons as a threat comparable to the one that led to the September 11, 2001 attacks. The document was announced on the eve of the 25th anniversary of the attacks, during a hearing held at the Flight 93 National Memorial in Shanksville, Pennsylvania, the site where the plane crashed after its passengers thwarted an attack on Washington.

What the committee found

The report's authors write bluntly that the United States "faces a threat environment as complex and dangerous as any since September 11, 2001," and that "serious gaps remain in our ability to confront these vulnerabilities." That assessment extends well beyond artificial intelligence alone, but AI occupies a central place in it as a factor reshaping the nature of both terrorist and technological threats at once.

Committee Chairman Rick Crawford stressed that the next major attack on the United States could look completely different from the one 25 years ago. The committee warns against repeating the mistake that preceded the 2001 attacks, namely a failure of imagination in anticipating new forms of threat.

The next strategic attack on the United States may not involve hijacked planes - Rick Crawford, Chairman of the House Intelligence Committee

Language models as a tool for terrorists

The most concrete part of the report concerns advanced language models. The committee writes that "the latest generation of language models could significantly ease the ability of any actor, including terrorist organizations, to develop and carry out more destructive attacks." That framing places large AI models in the same category as biological and cyber weapons, dual-use technology that can be exploited both defensively and offensively.

Congressman Josh Gottheimer, who co-led work on this section of the report, pointed to Iran, China and Russia as actors already using artificial intelligence today to strengthen intelligence and disinformation operations.

AI supercharges all of this. A single adversary can now generate fake voices and fake videos - Josh Gottheimer, member of Congress, House Intelligence Committee

Cyberattacks, drones and critical infrastructure

The report also cites the AI-driven acceleration of cyberattacks, noting that artificial intelligence sharpens both offensive and defensive tools in cyberspace at the same time. OpenAI and more than a hundred other companies previously warned that attacks aided by AI models are spreading on a scale of months rather than years. The committee also flags the risk of algorithm-controlled drone swarms capable of disrupting commercial airspace, as well as threats to telecommunications and power grids.

A separate thread concerns biotechnology. Former national security adviser H.R. McMaster warned the committee about the risk of combining artificial intelligence with biological research, which makes it easier to design dangerous pathogens with a lower threshold of specialized knowledge than before. The report also notes that China and Russia are deliberately escalating public debates around AI data centers in the US, exploiting genuine public concerns, such as energy consumption, for broader disinformation purposes.

Recommendations for intelligence and Congress

Rather than curbing AI development, the committee recommends the opposite direction. The document calls on the US intelligence community to accelerate its own responsible deployment of advanced AI capabilities, arguing that US adversaries are already doing so and that maintaining a technological edge is a condition for security, not a threat in itself.

Among the remaining recommendations out of the twenty are the implementation of a biometric entry-exit control system modeled on European Union solutions, building interoperable information-sharing systems between intelligence agencies and state, local and tribal authorities, reauthorization of Section 702 of the Foreign Intelligence Surveillance Act (FISA), and mandatory training for members of Congress on handling classified materials.

What this means beyond the United States

The US intelligence committee's report carries no binding force for other countries, but it sets the tone for a debate that European regulators and allied agencies later draw on. For Poland, which is only just building its own committee to oversee implementation of the EU AI Act, the American assessment of language models as a risk tool for non-state actors is a signal that AI-related national security issues extend far beyond market regulation and touch directly on defense and intelligence.

The report's publication coincides with a growing number of warnings from other institutions, from central banks to non-governmental organizations, that the pace of safeguards around AI systems is failing to keep up with the pace of their development. By issuing its warning at a symbolic place and on a symbolic day, the US intelligence committee signals that it treats this risk on par with classic terrorist threats rather than as a distant technological scenario.