Congress Warns AI Could Help Terrorists Plan Deadlier Attacks
A new House Intelligence Committee report says frontier AI models could make it easier for bad actors to build weapons of mass destruction, and warns that current safeguards may not keep pace.
Key points
- The House Permanent Select Committee on Intelligence published a report in late August 2025 warning that advanced AI poses serious national-security risks.
- The report names large language models, the AI technology behind tools like ChatGPT, as potentially lowering the barrier for terrorists to develop weapons of mass destruction.
- Bipartisan authors include House Intelligence Chair Rick Crawford and ranking member Jim Himes, alongside Reps. Elise Stefanik and Josh Gottheimer.
- The committee says the U.S. intelligence community is not adequately prepared to confront these new AI-enabled vulnerabilities.
- Lawmakers are calling for faster, responsible AI adoption inside spy agencies, paired with human oversight and civil-liberties protections.
The report landed less than two weeks before the 25th anniversary of the September 11 attacks. Its timing was deliberate. Congress wanted a hard look at where U.S. intelligence stands today, and what it found was unsettling.
"We are not adequately prepared to confront and address our vulnerabilities," the report states plainly.
What exactly is the AI threat?
The committee's core worry is that frontier large language models, the AI technology that powers chatbots and research assistants, could hand dangerous knowledge to people who would previously have struggled to find or use it. Think of it as a very capable, always-available research assistant with no ethical boundaries, if the safety guardrails built into it can be bypassed.
The report warns that while AI labs do try to stop their models from giving out instructions for, say, building a bioweapon, "the underlying capability of models is developing rapidly and such a use of a model is difficult to fully control." In plain terms: the locks are improving, but the door keeps getting bigger.
The committee specifically flags what it calls "Black Swan" risks. A Black Swan event is a rare, hard-to-predict catastrophe with massive consequences. The argument here is that AI lowers the skill bar for anyone wanting to cause one.
Should ordinary people be worried?
This report is aimed at policymakers, not the public. You are not being asked to act on anything today. But it does reflect a genuine and growing debate in Washington about how to regulate AI without slowing down the enormous economic benefits it is delivering.
First reported by CNBC Tech, the story sits at the intersection of two things Congress is struggling to balance at once: not letting the U.S. fall behind China or other rivals in AI capability, while also keeping the most dangerous applications out of the wrong hands.
What are lawmakers asking the government to do?
The committee's recommendations fall into two buckets. First, the intelligence community should adopt AI tools faster for its own collection and analysis work, so the U.S. keeps an edge over adversaries. Second, any tools used internally must come with rigorous testing, human oversight, and strong privacy protections.
The four lawmakers who authored the report are bipartisan, which matters. Agreement across party lines on AI security is still rare enough to be notable.
| Author | Party | State |
|---|---|---|
| Rick Crawford (Chair) | Republican | Arkansas |
| Jim Himes (Ranking Member) | Democrat | Connecticut |
| Elise Stefanik | Republican | New York |
| Josh Gottheimer | Democrat | New Jersey |
Common questions
What is a large language model?
A large language model is the AI technology behind chatbots like ChatGPT and Claude. It is trained on huge amounts of text and can answer questions, write documents, or explain complex subjects in seconds.
Can AI really help someone build a weapon?
Current AI tools have safety filters designed to block dangerous instructions. The committee's concern is that those filters are imperfect and that the underlying capability of the models is advancing faster than the safeguards around them.
Does this mean AI will be regulated soon?
Congress is actively debating AI regulation, but no federal law has passed yet. This report adds pressure, though it stops short of proposing specific legislation.
