Fable JUST made EVERYONE MAD...
Credibility score: 45/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "Fable JUST made EVERYONE MAD..." and rated it 45/100 for credibility (a BS score of 55/100 — mixed credibility), on 2026-06-11. Its weakest claim — "Big tech and banks got Mythos via Project Glass Wing" — scored 30/100 and was flagged as sketchy. 17 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 17 claims analyzed: 6 scored under 40, 11 between 40 and 69, and 0 at 70 or above.
Claims analyzed
Anthropic's model silently sabotages users trying to build their own AI — Dubious (45/100)
At 0:30
They published the exact behaviors in the system card — that's the opposite of 'silent.'
Why this score: Key contradiction in wording.
- The transcript itself later admits the safeguards are documented and users get steered to weaker models
- Anthropic publicly listed the restricted areas (biology, cyber, distillation)
- Calling documented, announced behavior 'silent sabotage' is the part people are debating
Original quote: “Enthropic will be silently preventing people from using these models to develop their own AI models. In fact, most of the things that have to do with machine learning, AI, etc. This model will silently sabotage you”
Claude is already silently blocking GPU inference research — Sketchy (35/100)
At 2:30
SemiAnalysis claim with zero public logs or reproducible examples shown.
Why this score: The assertion relies on an unnamed "semi analysis" source and anecdotal reports. No independent verification or logs have surfaced to confirm systematic blocking of standard GPU inference queries.
Original quote: “We already are seeing Anthropic's latest models moderation filters are GPU inference research and programming.”
AI companies are silently steering models to push agendas — Opinion (50/100)
At 4:30
Calls out the exact fear — models changing behavior on the fly to steer users.
Why this score: Core concern is opaque steering by AI firms.
- No public evidence yet of Anthropic doing this at scale.
- The "contract" he mentions was never a formal agreement, just industry norms that are already eroding.
Original quote: “>> That is the real challenge. How are people going to know when it's being steered or not? And thankfully, there's a little bit of work around this that Anthropic themselves have done, but there's a lot more to be done. And I think it's going to kick up a lot now”
AI steering is exactly like Three-Body Problem aliens sabotaging science — Dubious (35/100)
At 4:50
Three-Body Problem aliens are fictional. Real AI bias is messy, not sci-fi sabotage.
Why this score: Analogy is dramatic.
- No evidence of coordinated scientific-result manipulation by AI labs.
- Current issues are more about training data and safety policies than alien-level conspiracy.
Original quote: “They're already showing here. We can do it and we will do it for reasons that we justify only to ourselves. I'm sure everyone here also is probably familiar with, you know, the threebody problem. This is literally like the exact story in the threebody problem with the sans where the aliens like…”
Claims AI control theory is established fact everyone in open community already knows — Opinion (50/100)
At 6:30
Calling it settled theory when it's still mostly speculation — big leap.
Why this score: Opinion framed as fact.
- No specific evidence or studies cited for the "theory"
- Relies on authority of unnamed "open community"
- Regular people not being "outraged" doesn't prove the theory is correct
Original quote: “the theory is real and what all of us in the open community have been saying for years and yet regular folks don't get it yet”
Predicts AI will soon punish users for wrong opinions by blocking or reporting them — Dubious (35/100)
At 6:37
Jumping from current guardrails to full thought police — that's quite a leap.
Why this score: Speculative escalation.
- Current AI refusals exist but are narrow (harm, illegal activity)
- No evidence of widespread "wrong opinion" reporting systems
- Conflates content filters with active punishment
Original quote: “tomorrow it'll be I can't do this with my computer because the AI stopped me or it reported you because you said something it didn't agree with. You had the wrong opinion and must be punished for it”
Claims Anthropic refused Pentagon deal over spying on citizens and autonomous weapons — OK (60/100)
At 6:50
The red lines part checks out publicly — the rest is a bit fuzzy on details.
Why this score: Partially supported.
- Anthropic did publicly state concerns about autonomous weapons and citizen surveillance
- Exact contract details and "kicked out" framing aren't fully confirmed in public records
- Government still uses their models, which matches the transcript
Original quote: “Pentagon wanted in the contract, they wanted an agreement that they could use Claude for all lawful purposes and they kind of buted heads with Enthropic because Dario and Enthropic they were saying there their two red lines were not willing to cross. One is spying on US citizens and two is use of…”
Claims Pentagon listed Anthropic as supply chain risk after contract issues — Unverifiable (50/100)
At 7:10
No public record of that exact designation — sounds like insider knowledge.
Why this score: No supporting evidence found.
- No official Pentagon statements or reports confirm Anthropic as supply chain risk
- Public records show continued government use of their models
- Could be accurate but currently unverifiable from open sources
Original quote: “the Pentagon did list Anthropic as a supply chain risk. It doesn't seem like it affected them that much”
Fable silently censors via prompt tweaks, never tells users — Dubious (45/100)
At 8:30
No receipts on the vector/PFT sabotage method — sounds like speculation dressed as insider knowledge.
Why this score: No public evidence of Fable using hidden prompt modifications. Key issues:
- Speaker describes specific technical sabotage without naming sources
- Recent Fable coverage focuses on gameplay demos, not backend censorship
- Community feedback shows frustration with refusals, not proof of stealth downgrades
Original quote: “fable won't fall back to a different model and tell you it just limits the output through prompt modifications during vectors or PFT like we talked about you won't be told when it happens to you”
AI labs are openly admitting quiet user sabotage is fine — Opinion (50/100)
At 8:42
Calling it 'they're saying it's okay' without a quote — that's the speaker's framing, not a direct admission.
Why this score: This is interpretive language rather than a sourced claim. The transcript shows the speaker labeling the behavior, not citing any lab statement that endorses 'quiet sabotage.'
Original quote: “They're saying it's okay for us as the AI labs to quietly sabotage the user and steer them in a different direction because we feel, you know, that's the correct thing to do. And I do believe silent sabotage is the correct word here”
Future AI will secretly steer users politically — Opinion (50/100)
At 8:56
Classic slippery slope — possible in theory, zero evidence it's happening or planned right now.
Why this score: Speculative future risk. No current documentation shows AI labs embedding political steering at scale. The claim jumps from today's content filters to deliberate worldview engineering without intermediate proof.
Original quote: “In the future, it might be used to push you in a particular political direction or to shape the world closer towards somebody's sort of ideal image”
OpenAI leaked memo from last 24 hours links RSI to delaying IPO — Dubious (35/100)
At 9:10
No link, screenshot, or source provided for this 'leaked memo' — just the speaker reading it aloud.
Why this score: No verifiable source for the memo. Key issues:
- Transcript gives no link, date, or author beyond 'Alman'
- Web search returns zero results for any OpenAI memo from June 10-11 2026 matching this text
- OpenAI IPO plans have been discussed publicly for months; nothing new surfaced in last 24 hours per available data
Original quote: “Here's a little leaked memo from inside of OpenAI, for example, just from the last 24 hours. Alman said that if the company's technology advances at a rapid pace to the point where the AI itself is able to create new AI, known as recursive self-improvement, that would lessen the chance of a quicker…”
OpenAI is definitely IPOing in 2026 — OK (60/100)
At 9:30
OpenAI has signaled IPO interest, but 'planning to IPO this year' is still an unconfirmed timeline.
Why this score: OpenAI leadership has publicly discussed going public eventually. No firm 2026 date has been locked in as of June 2026. The speaker treats intent as a done deal.
Original quote: “Openai is planning to IPO this year. They're planning to take their company public on the on the stock exchange, right? So, it's kind of a short-term financial plan to do this thing, right? So, we're halfway through the year. We got about 6 months left”
Says Sam Altman is weighing RSI takeoff speed for OpenAI's IPO decision — Dubious (35/100)
At 10:30
No public statement from Altman links IPO timing to any 'RSI takeoff' metric.
Why this score: No evidence supports this specific claim.
- Altman has discussed IPO considerations publicly, but no statements reference RSI speed as a factor.
- 'RSI takeoff' appears to be the speaker's invented framing, not something Altman has cited.
- This conflates general IPO strategy talk with a specific technical threshold that hasn't been documented.
Original quote: “Sam Alman is taking into consideration how fast the RSI takeoff will be to make that decision.”
AI-powered tyranny could be the last regime in history — Opinion (50/100)
At 12:30
Calling it the final tyranny ever feels like dramatic escalation, not analysis.
Why this score: Speculative framing rather than evidence-based. Key issues:
- No historical or technical data cited for permanent regime claim
- Treats AI as uniquely unstoppable compared to past tech
- Speaker explicitly labels it a personal agreement, not a researched position
Original quote: “his warnings of this idea of AI powered suppression, of AI powered tyranny, which which he's described in multiple blog posts, and I agree with everything that he's saying. It is a very strong threat. We should be incredibly careful. An AI powered tyrant might be the last one in history.”
Big tech and banks got Mythos via Project Glass Wing — Sketchy (30/100)
At 14:30
Project Glass Wing also missing from any records — this tier list feels made up on the spot.
Why this score: Zero sourcing for either project name.
- No announcements, leaks, or filings mention Glass Wing
- Anthropic's public models (Claude 3.5/4) show standard tier access, not secret bank deals
- Claim relies on insider knowledge that isn't verifiable
Original quote: “The biggest banks who has access to mythos, right? The people in project Glass Wing. We have first and foremost the biggest tech companies, right? Apple Google Microsoft Crowdstrike AWS, Nvidia.”
Anthropic silently routes biology/cybersecurity prompts to weaker models — Dubious (45/100)
At 16:30
Big accusation with zero receipts — just stated like it's established fact.
Why this score: No public evidence presented for silent model routing on specific topics.
- Speaker offers no logs, benchmarks, or leaked internal docs.
- This would be a major safety/ethics issue if true, yet no corroborating reports exist.
- Relies entirely on the speaker's assertion without supporting data.
Original quote: “the model is degraded silently without you knowing or in the case of biology and cyber security routed to a lower model.”
See the full analysis with timestamps →