The First Domino Has Fallen...
Credibility score: 40/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "The First Domino Has Fallen..." and rated it 40/100 for credibility (a BS score of 60/100 — mixed credibility), on 2026-06-16. Its weakest claim — "Mythos precedent means only gov + Anthropic will keep cyber research access" — scored 25/100 and was flagged as sketchy. 7 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 7 claims analyzed: 3 scored under 40, 4 between 40 and 69, and 0 at 70 or above.
Claims analyzed
US gov just banned Mythos/Fable AI models worldwide via export controls — Sketchy (35/100)
At 0:00
Claims a federal export ban on Mythos/Fable this weekend — zero evidence of any such order existing 💀
Why this score: No matching government action found in current records.
- No export control announcement targeting Mythos or Fable appears in official sources
- Anthropic has not publicly confirmed or acted on any worldwide ban
- Community comments treat this as rumor or exaggeration rather than confirmed policy
Original quote: “The first domino has fallen in what I think will be a series of unfortunate events as it applies to not only the use of AI, the democratization of AI, but the world of AI enabled cyber security research and cyber security as a whole. What I'm talking about is the ban this weekend of mythos, right?…”
Amazon researchers jailbroke Fable into public Mythos — Sketchy (30/100)
At 0:30
Amazon found a Fable jailbreak turning it into Mythos — still zero receipts on this story 🚩
Why this score: No evidence of Amazon researchers publishing or leaking such a jailbreak.
- No reports, papers, or leaks mention this specific bypass
- Speaker presents it as fact without sourcing
- Sounds like dramatic retelling of unverified rumors
Original quote: “Apparently researchers at Amazon found a jailbreak in Fable that allowed you to use Fable the model with guard rails to do security research. You know, bypassing all the rules that Anthropic put around it and therefore making it effectively a publicly available mythosclass model.”
No universal jailbreak found that unlocks broad cyber capabilities in Mythos 5 — Dubious (40/100)
At 3:01
Zero proof offered — just 'testers haven't found one yet.' That's not evidence, that's absence of evidence.
Why this score: No testing data or researcher names provided. Key issues:
- Relies on 'testers' without naming who or what was tested
- The Amazon jailbreak is mentioned as limited but no benchmark numbers given
- Public red-teaming results for Mythos 5 remain scarce
Original quote: “no testers have yet to have been yet been able to find a universal jailbreak, a jailbreak method that can very broadly bypass the model safeguards, unblocking a wide variety of cyber capabilities.”
US gov banned foreign access to Fable 5 & Mythos 5 over national security — OK (55/100)
At 3:50
Sounds dramatic — no actual directive text shown, just speaker's recap. Could be real or PR spin.
Why this score: No primary source cited. Key issues:
- Speaker paraphrases the order without linking the actual document
- Anthropic's own blog post is the only evidence referenced
- Timing and scope match recent AI export-control chatter but remain unverified in public records
Original quote: “The US government citing national security authorities has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national whether inside or outside the US including foreign national anthropic employees.”
Other models beat Mythos on cyber benchmarks without needing the banned model — Dubious (45/100)
At 3:58
Drops 'Cyberjim.io' and 'Mdash' like we should know them — zero links, zero numbers. Feels like flexing with mystery tools.
Why this score: Benchmarks and tool names are unverifiable here. Key issues:
- No actual leaderboard screenshots or paper citations
- Personal zero-day claims can't be checked
- Microsoft Mdash reference appears nowhere in public cyber-AI literature
Original quote: “you don't need Mythos to do cyber security research. I myself have found zero days using Opus 47 with Gedra MCP and a couple cool harnesses that I wrote. Even there's proof publicly that there are models when used in the proper harness like Microsoft's Mdash for example that don't use Mythos and…”
Mythos precedent means only gov + Anthropic will keep cyber research access — Sketchy (25/100)
At 4:30
One public model getting restricted = total lockdown for everyone else? Massive leap, zero evidence 🚩
Why this score: Slippery slope fallacy.
- Assumes one lab's decision instantly becomes industry-wide policy
- Ignores open-weight models and competitors still shipping research access
- *'Could lead to'* is doing all the heavy lifting
Original quote: “Maybe these companies just don't give us access to cyber research period. Maybe maybe only the government and anthropic can do cyber security research in these models. I'm not saying that's going to happen. I'm just saying the precedent we're setting that mythos is dangerous. Cut it off from the…”
Says you must actively hack with Metasploit and Ghidra to understand exploits — Opinion (50/100)
At 6:30
Bro really said the ONLY way to learn hacking is committing crimes in a VM 💀
Why this score: Opinion on learning method.
- Speaker presents personal view that hands-on illegal activity is required
- No evidence cited that this is the sole effective approach
- Alternative legal resources (CTFs, labs) exist and are widely used
Original quote: “you need to know how hackers work. You need to know how exploitation occurs. And the only way you can do that is by doing it, right? By using tools like Metas-Loit, by going into GEDra and ripping apart a binary.”
See the full analysis with timestamps →