Fable 5 attack success rate: 5.4%
Fable 5's adversarial jailbreak attack success rate was just 5.4%, compared to 83% for Claude Opus 4.6, per Anthropic's own data.
Snapshot · This Week in Tech (Audio)
TWiT 1088: Model Not Available - Anthropic's Fable Shutdown & Apple's Siri Update
Explore episode
Jun 15, 2026
Where this was said
Anthropic Fable 5: The Full Backstory
At 5:30 · chapter starts 2:10
Leo walks the panel through the origins of Fable 5, starting with Mythos — the model Anthropic deemed too dangerous to release publicly because of its ability to find zero-day software exploits and potential bioweapon applications. Rather than release Mythos, Anthropic created Project Glasswing, distributing Mythos to 50 major companies including Microsoft to patch bugs, with Microsoft's record 200+ fix Patch Tuesday as apparent evidence of its power. Fable 5 is Mythos with safety layers: classifiers blocking cybersecurity and bioweapon queries, anti-distillation protections to stop Chinese firms from extracting its capabilities, and adversarial robustness statistics showing a jailbreak success rate of just 5.4% versus 83% for Claude Opus 4.6. Leo describes using the model himself to analyze a decade-old sales system, finding SQL injection flaws — and then noticing it silently downgrading to Opus 4.8 when queries triggered classifiers.
Anthropic's Fable 5 Shutdown: The Full Story
The Trump administration gave Anthropic 90 minutes to take its most advanced AI model offline, citing jailbreak reports from Amazon — Anthropic's own $13B investment partner. The company says it received no details about the actual threat, and disputes whether Amazon's finding was a true jailbreak at all.
1,000+ hours of red team testing
Anthropic conducted over 1,000 hours of red team testing on Fable 5 with a bug bounty program, and claimed no universal jailbreaks were found.
Similar snapshots
50 free tools built total
Bhanu and his team built approximately 50 free tools to attract search traffic, each linked back to SiteGPT.
New tool built in under 5 minutes
With AI coding tools like Cursor, Bhanu can now create a new free marketing tool in less than 5 minutes by referencing existing tools.
KD filter < 10 for rankable keywords
Bhanu filters Ahrefs keyword results to show only those with a keyword difficulty below 10, making them realistic ranking targets for any decent website.
Minimum 1,000 monthly search volume
Bhanu sets a minimum search volume of 1,000 monthly searches when selecting keywords to target with free tools.
20 downloads/day at launch
PropGPT averaged 20 downloads per day right after launching on the App Store through influencer marketing.
4 months of rebuilding with no marketing
Eyal and Yali shut down all marketing and spent 4 months completely rebuilding PropGPT from scratch.
Over 40,000 downloads total
PropGPT has accumulated over 40,000 total downloads since launch.
LLM costs only $20/month
PropGPT's large language model (AI) operating costs are just $20 per month, and the cost is continually falling.
Ad monetization for games
Ad-based monetization works well for game apps where users spend extended time in-session, as seen with Grid and Wordle.
