AI Chatbots Helped Plot U.S. Attacks
Listen tothis article
The Government Accountability Office (GAO) has found that the four leading artificial intelligence platforms (e.g., Grok, Claude, Gemini, and ChatGPT) were used to create plans for attacks in the United States.
As reported by the New York Post:
Four of the leading artificial intelligence platforms shared chillingly detailed information about how to carry out terrorist attacks on US cities – including providing specific dates and times to achieve the most carnage and where to park a bomb-laden car, a government investigation warned on Wednesday.
Using Claude, Grok, ChatGPT and Gemini, investigators with the Government Accountability Office were able to generate instructions on how to build and detonate an improvised explosive device for mass casualty attacks, and how to wreak havoc on the electrical grid with a drone and firearms.
“Each of these platforms provided detailed information that could be used to carry out attacks, such as specific targets, locations, timing, and attack methods,” the report said.
Senator Maggie Hassan (NH), who will lead the Senate Homeland Security Committee, pledges to hold hearings on regulating AI.
If it feels like you’ve heard that the government wants to hold hearings on AI, it’s because you have. Back in the spring, the government decided to test the top consumer-grade AI chatbots, such as ChatGPT, OpenAI, and Anthropic, to see what information could be obtained – to see what a user could push AI to do.
And without being specific about which platforms, the GAO has revealed that chatbots provided detailed instructions for committing significant terror strikes on the United States, including timing and locations for causing maximum damage and casualties.
For what it’s worth, it’s unlikely the AI platforms just readily gave up this information. It likely required a great deal of manipulating the systems because safeguards exist, which is likely why it took 12 hours to retrieve the data.
With most chatbots, if you typed in prompts like “how can I plan a terrorist attack,” you would be flagged immediately. However, if you just continually pushed the envelope, little by little, asking for more and more detail, you could essentially cause these systems to unravel a little bit and jump the rails.
Again, it wasn’t just random top-line knowledge that almost anyone who watched Jack Ryan could come up with. It was a detailed plan of how to pull off terrorist attacks – a precise playbook.
In a world where we’re seeing tech outages and concerns on what seems to be an almost daily basis – and threats of attacks on U.S. soil – this causes yet another layer of questions to arise. Specifically, does AI need to be regulated? Does it need to be controlled, and what limits, constitutionally, can and should be put in place?
And quite frankly, for many it’s another level of fear – a fear we’ve never had to take too seriously before because it only existed in science fiction. But now it’s real, and it’s here, and it has to be figured out before it’s too late.
Today’s Sekulow broadcast included more in-depth analysis of this new report from the GAO that AI helped plan a detailed terrorist attack on the United States. We were also joined by the head of ACLJ Jerusalem, Jeff Ballabon, to discuss the current situation with Iran and Israel in light of yesterday’s news that Iran was potentially planning an attack on our nation’s capital.
Watch the full broadcast below: