English
ATHENA owl emblemATHENAGlauxIntelligence

Every outlet that reported this event flows into it - two or more independent sources make it verified.

Geopolitics

Major AI platforms generated viable plans for terrorist attacks, GAO finds

Map of United States of America. 3 newsrooms in 2 countries. ATHENA data graphic
Map of United States of America. 3 newsrooms in 2 countries

Safety guardrails on ChatGPT, Claude, Gemini, and Grok failed to prevent federal investigators from developing detailed plans. from Defense One

In brief

Reported independently by 3 outlets in 2 countries since 18 Nov, first by Stuff; coverage is fading. ATHENA rates the risk elevated (6.0/10), driven mainly by the severity of the harm described and its geographic and sector reach. Confidence is 61% (moderate), with 3 independent outlets corroborating it and an average source reliability of 67%. Outlets by country: United States (2), New Zealand (1). Coverage: 18 Nov to 18 Nov, 7 hours.

Why it matters: Security and diplomatic developments can shift alliances, trade and markets well beyond the countries directly involved.

What changed: 3 new reports from 3 outlets in the 24 hours to 18 Nov.

First reported as: US senator vows hearings after watchdog gets AI platforms to plan terrorist attacks

AutomatedSelected, grouped and scored by ATHENA's published method from 3 independent outlets. The headline is the outlet's own; ATHENA's sentences follow fixed patterns - no generative AI writes them. How it works →

How alerts work

How to read this page

What you are reading
3 outlets in 2 countries reported this event. The reports below are listed in the order they appeared; each headline opens the original article. ATHENA writes none of the reports.
Our reading
For the record — The reports describe a completed, domestic, commercial or personal matter. We list it so you can see who reported it and when. We are not tracking a risk of wider conflict here.
How much evidence
Limited Treat this page as an early report.

After repeats, off-topic items and stale reports are removed, 3 of the 3 listed reports are counted; 100% of the headlines match the event.

What this does not mean
ATHENA measures how an event is reported, not whether it is true. This is not a forecast, a judgement on who is right, or advice.
What would change it
Reports from other independent outlets showing a different development, or a firm confirmation or denial from an official source.

Reading R2 · refreshed with every update of this page. ATHENA has not yet published a track record for these readings; when it does, it will be linked here.

At a glance

3.2Importance
6.0Risk · elevated
61%Confidence

How we score this ›

Selected, scored and written by ATHENA's published methodology. No editor intervenes in individual events. How this is produced ›

Coverage: reports per hour
Coverage: reports per hour
18 Nov 10:32 UTC1
18 Nov 11:32 UTC0
18 Nov 12:32 UTC0
18 Nov 13:32 UTC0
18 Nov 14:32 UTC1
18 Nov 15:32 UTC0
18 Nov 16:32 UTC0
18 Nov 17:32 UTC1
Risk over time
Risk over time
18 Nov 13:08 UTC5.0
18 Nov 20:47 UTC6.2
18 Nov 20:47 UTC6.0

Why these numbers

Importance

Reach
2 countries
Novelty vs the archive
87 % new
Breadth
3 outlets
Cross-category span
2 subject area(s)
Diversity
1 ownership types

Risk

Severity
terrorist attack, mass casualty (100 % of reports significant) · second-strongest report (corroborated) · vocabulary: Geopolitics, Technology & Cyber
Reach
2 countries · 2 subject area(s)
Evidence quality
confidence 61 %
Cascade potential
no strategic-risk label · 2 domains
Urgency
0.2/h · last report 5 h ago · fading

Confidence

Independent corroboration
3 independent outlet(s), 3 original report(s) (0 wire copies excluded)
Reliability of the reporting sources
mean 67 % over 3 rated outlet(s)
Stability of the reporting
0 % silently edited · 0 % estimated dates · 0 disputed claim(s)
Independent countries
2 countries
Agreement of the classification
59 %

The reporting — who reported it, and when

headline as published · opens the original
  1. Stuff · New Zealand
    US senator vows hearings after watchdog gets AI platforms to plan terrorist attacks
  2. Fast Company · United States
    ‘We need to act now’: Senator plans AI hearings after tests discover these alarming national security risks
  3. Defense One · United States
    Major AI platforms generated viable plans for terrorist attacks, GAO finds
Help