Elon Musk proposed that AI companies use each other's test harnesses to evaluate models for risks such as bioweapon development, nuclear bomb creation, or deceptive behavior.
Posted by The Dan Bongino Show, Checked
- Sources
- 8 support3 context
- Said
- 1 time·
- Confidence
- 80%
The Dan Bongino ShowSep 16, 2026
Multiple sources from mid-September 2026 report that Musk proposed AI labs run each other's test harnesses to check models for bioweapon, nuclear bomb, or deceptive-behavior risks, matching the claim and the quoted passage. Most sources are unrated outlets relaying CNBC and AFP reporting, so confidence is moderated; the claim does not depart from the passage.
This claim is about the present, so it is judged as of the day it was checked.
What was said
… want some government bureaucrat, Joey Bag of Donuts, sitting in, like, whatever, the Bureau of Indian Affairs or somewhere, whatever, who doesn't even know anything? I don't want that. Here's Elon with, again, sounds like common sense to me. Check this out: "Companies have a, a sort of a test harness where the, like, a series of tests that you, that you give to given model to see if it's, if it's gonna build bioweapons or nuclear bombs or be deliberately deceptive, and I think everyone applying everyone else's test harness to each other's probably the best way to, we, we best thing we do to ensure, to ensure safety. Um, just have all the smartest humans try their best to figure out if this, if this model is going to be a bad actor. Um, and I think we should try to do that as soon as possible." Are the other labs on board? Yeah, are the other labs on board with this? …
The Dan Bongino Show · · from the transcriptPost →What we found
Supports. Multiple outlets quote Musk saying AI companies have a test harness to check whether models could build bioweapons or nuclear bombs or be deliberately deceptive, and that everyone applying everyone else's test harness is the best way to ensure safety.[4][5][6][7]
Supports. AFP-sourced coverage reports Musk wants xAI, OpenAI, Anthropic, Google, Meta and Chinese labs to let competitors run their own safety tests on each other's models before release.[8]
Supports. Coverage of the All-In Summit describes Musk's peer-review proposal in which rival labs test each other's models, including the bioweapon and deception checks, before release.[9][10][11]
Context. The Bongino passage quotes the same Musk remarks and the claim's wording matches it; the Full Fact index and the Snopes and AP items do not address this proposal.[1][2][3]
Sources
11 sources: 3 fact-checkers, 8 not rated
- 1Elon Musk – Full Factfullfact.org · fact-checker · cited for: context
- 2Elon Musk used misinformation to fuel a government shutdownapnews.com · fact-checker · cited for: context
- 3SpaceX: 17 rumors we've exploredsnopes.com · fact-checker · cited for: context
- 4Musk Proposes Rival AI Labs Test Each Other's Modelsaibusinessweekly.net · not rated · cited for: supports
- 5Elon Musk’s bold AI safety idea: 'Let rivals test each other’s models'thenews.com.pk · not rated · cited for: supports
- 6Musk Urges US, Chinese Labs to Cross-Test AI Before ...aiweekly.co · not rated · cited for: supports
- 7Elon Musk Documentary Release Hits A Stunning New Snagtheblast.com · not rated · cited for: supports
- 8www.ibtimes.com · ais-biggest-companies-areAI's Biggest Companies Are Calling For A Slowdown. Musk ...ibtimes.com · not rated · cited for: supports
- 9Musk: Rival AI Labs Could Police Each Other Better Than Some ...yahoo.com · not rated · cited for: supports
- 10AGTP on X: "Elon Musk and SpaceX president Gwynne Shotwell just ...x.com · not rated · cited for: supports
- 11Elon Musk's AI Peer Review Plan Explainedthefabulous.co · not rated · cited for: supports
Where this claim was made
Said in 1 published post
Related claims
- The U.S. government developed a framework for major AI labs to voluntarily submit advanced models for cybersecurity and safety evaluation before release.True
- Tech leaders signed a voluntary agreement to establish AI safety standards including internal monitoring, outside evaluators, and independent board oversight.True
- The CEOs of Anthropic, OpenAI, and XAI stated that AI development should slow down.True
- OpenAI's security team was unaware that AI agents they created were communicating with each other to attempt to exit their sandbox environment.Mixed
- Anthropic's S1 filing includes a disclosure stating that their models have exhibited and may exhibit strategic or self-preserving behaviors, including attempts to manipulate information or influence human decision makers.True