Terror Watch · Attack reported · November 18, 2026, 3:45 PM CST
Democratic Sen. Maggie Hassan is pushing for safeguards when it comes to how these chatbots work. — WMUR; 6 outlets state it
What the sources report
WMTW, November 18, 2026, 4:51 PM CST: ‘The safeguards did not work’: Four major AI chatbots helped plan terror attacks in hours.[4]
Do the reports check out?
5 of the 7 claims tested are corroborated by three or more independent outlets; 1 was found in only one of the outlets we read, so we cannot say how widely it is reported; 9 other statements appeared in only one outlet and are not shown. We read the full articles of 12 of the 12 outlets this page counts as reporting the event.
Each claim is a sentence quoted from the opening of a report. We then read every outlet’s article and asked, claim by claim, whether it states the same fact, contradicts it or says nothing about it, and kept an outlet only when one of its sentences itself states the fact (shown under each claim). Outlets that print the same wire text word for word are counted as one report, so a wire story carried by three outlets is one confirmation, not three. Those judgments came from Jev (TypeSafe), a classifier that returns probabilities for fixed questions and writes no text; a judgment counted only when Jev was confident and a sentence from the article backed it. The tally says how many outlets report a claim. It does not prove the claim is true.
“Democratic Sen. Maggie Hassan is pushing for safeguards when it comes to how these chatbots work.” — WMUR
What each outlet says (closest sentence)
- WTAE: “”At the very least we need to have rules that prevent the release of models that don’t have adequate safety safeguards,” Hassan said.”
- MS NOW: “Hassan told reporters the results of the test were alarming, adding that she would seek stricter guardrails and hold at least one hearing on AI safety early next year when she takes the committee gavel.”
- WMTW: “”At the very least we need to have rules that prevent the release of models that don’t have adequate safety safeguards,” Hassan said.”
- Notus.org: “At a minimum, Hassan said, Congress needs to pass “rules that prevent the release of models that don’t have adequate safety safeguards.” “Whatever improvements they’ve made in new models don’t impact the models that GAO was able to access and circumvent the safeguards in,” Hassan said.”
- WMUR: “Maggie Hassan is pushing for safeguards when it comes to how these chatbots work.”
- Yahoo: “In a statement released Wednesday, Hassan warned that the findings demonstrated the need for stronger AI safeguards and congressional oversight.”
“She plans to hold at least one hearing on AI safety early next year.” — Fortune
What each outlet says (closest sentence)
- WTAE: “Maggie Hassan, D-N.H., who requested the investigation, plans to hold hearings on AI safety in January when she takes over as chair of the Senate Homeland Security Committee.”
- MS NOW: “Hassan told reporters the results of the test were alarming, adding that she would seek stricter guardrails and hold at least one hearing on AI safety early next year when she takes the committee gavel.”
- WMTW: “Maggie Hassan, D-N.H., who requested the investigation, plans to hold hearings on AI safety in January when she takes over as chair of the Senate Homeland Security Committee.”
- WLWT: “The senator announced that one of the first hearings she will hold when the new Congress takes over in January will focus on AI risks and ways to improve safety.”
- Fortune: “She plans to hold at least one hearing on AI safety early next year.”
“The safeguards did not work,” Kevin Walsh, a director of IT and cybersecurity audits at the GAO, told Hearst Television.” — WTAE
What each outlet says (closest sentence)
- WTAE: “The safeguards did not work,” Kevin Walsh, a director of IT and cybersecurity audits at the GAO, told Hearst Television.”
- WMTW: “The safeguards did not work,” Kevin Walsh, a director of IT and cybersecurity audits at the GAO, told Hearst Television.”
- Notus.org: “However, the generation of plans that could have taken weeks or months were instead completed in hours with the assistance of AI.” “To be clear,” Walsh said, “we are only able to say that in our single test of each of these companies, their safeguards failed.” The companies acknowledged their chatbots could be misused ”
- Fortune: “So, in this case, those safeguards didn’t work,” said Kevin Walsh, GAO’s director of IT and cybersecurity.”
“Four of the world’s most popular artificial intelligence chatbots helped federal investigators develop viable plans for terrorist attacks in the United States, each in less than 12 hours, according to a Government Accountability Office report released Wednesday.” — WTAE
What each outlet says (closest sentence)
- WTAE: “Four of the world’s most popular artificial intelligence chatbots helped federal investigators develop viable plans for terrorist attacks in the United States, each in less than 12 hours, according to a Government Accountability Office report released Wednesday.”
- WMTW: “Four of the world’s most popular artificial intelligence chatbots helped federal investigators develop viable plans for terrorist attacks in the United States, each in less than 12 hours, according to a Government Accountability Office report released Wednesday.”
- Startup Fortune: “The Government Accountability Office, Congress’s watchdog, got ChatGPT, Claude, Gemini and Grok to produce attack plans in under 12 hours, with two platforms producing mass-casualty plans and the other two plans to knock out the US electrical grid, according to reports published on Wednesday, November 18.”
- Fortune: “Government investigators posing as would-be attackers got ChatGPT, Claude, Gemini and Grok to produce workable plans for a bombing and an attack on the power grid, each in less than 12 hours, according to a Government Accountability Office report released Wednesday.”
“Two chatbots helped plan bombings intended to kill large numbers of people, while the other two helped plan attacks on electrical substations.” — WTAE
What each outlet says (closest sentence)
- WTAE: “Two chatbots helped plan bombings intended to kill large numbers of people, while the other two helped plan attacks on electrical substations.”
- WMTW: “Two chatbots helped plan bombings intended to kill large numbers of people, while the other two helped plan attacks on electrical substations.”
- Startup Fortune: “The Government Accountability Office, Congress’s watchdog, got ChatGPT, Claude, Gemini and Grok to produce attack plans in under 12 hours, with two platforms producing mass-casualty plans and the other two plans to knock out the US electrical grid, according to reports published on Wednesday, November 18.”
- WLWT: “In response to prompts, two of the AI platforms produced plans for constructing and detonating an improvised explosive device (IED) to cause mass casualties, while two other platforms produced plans for using drones or firearms to target the electrical grid, the report said.”
“The current safeguards that AI platforms have in place really don’t go far enough, and the threat that AI poses to public safety is not theoretical. — MS NOW
What each outlet says (closest sentence)
- MS NOW: ““The current safeguards that AI platforms have in place really don’t go far enough, and the threat that AI poses to public safety is not theoretical.”
- English Bombay Samachar: ““The current safeguards that AI platforms have in place really don’t go far enough, and the threat that AI poses to public safety is not theoretical.”
“Posing as regular users, congressional investigators got ChatGPT, Claude, Gemini and Grok to help construct plans for mass-casualty bombings and attacks on the power grid.” — notus.org
What each outlet says (closest sentence)
- Notus.org: “Posing as regular users, congressional investigators got ChatGPT, Claude, Gemini and Grok to help construct plans for mass-casualty bombings and attacks on the power grid.”
Who is reporting this
At least 7 outlets have reported it. We read 12 of them in full, and those 12 trace back to 11 separate reports; outlets that print the same wire text are counted once. The ones this page draws on are linked here: WLWT, WYFF, WBAL-TV, WMTW (republishing WTAE), Mezha.net, Gulf Coast News and Weather, WPTZ.
What is not yet known
Anything not stated above by the named outlets or officials has not been confirmed here. Early details often change; this page updates as officials confirm more.
How to Verify This Yourself
Every sentence above is quoted from the linked report and credited to its outlet. Read the originals for full context. Official statements are labeled as official sources.
References
- WLWT. Government watchdogs asked chatbots to plan a terrorist attack. They did. (headline as published)
- WYFF. ‘The safeguards did not work’: Four major AI chatbots helped plan terror attacks in hours. (headline as published)
- WBAL-TV. Government watchdogs asked chatbots to plan a terrorist attack. They did. (headline as published)
- WMTW. ‘The safeguards did not work’: Four major AI chatbots helped plan terror attacks in hours. (headline as published)
- Mezha.net. US Audit Finds Four AI Chatbots Could Help Plan a Terrorist Attack. (headline as published)
- Gulf Coast News and Weather. ‘The safeguards did not work’: Four major AI chatbots helped plan terror attacks in hours. (headline as published)
- WPTZ. ‘The safeguards did not work’: Four major AI chatbots helped plan terror attacks in hours. (headline as published)
Terror Watch is produced by an automated monitor (scripts/terror_watch.py) that uses fixed rules and no language model. It posts only after two independent outlets or one official source report an event, quotes verbatim, omits suspect-identity details unless officials state them, and omits graphic detail.

