Prof G Markets The Most Dangerous AI Model Just Leaked
186 snips
Apr 23, 2026 
Sherri Davidoff, LMG Security founder and cybersecurity specialist, breaks down the Anthropic Mythos leak, AI safety fears, patching chaos, and possible disclosure rules. Seth Goldstein, Morningstar equity analyst covering Tesla, digs into Tesla earnings, robotaxi progress, Optimus, and why autonomy keeps the valuation story alive. It also touches on how markets are reacting to the Iran war.
AI Snips
Chapters
Transcript
Episode notes
Mythos Changes Cybersecurity From Patch Lag To Instant Exploits
- Sherri Davidoff says Mythos rewrites cyber risk because it can find vulnerabilities and generate exploits even without source code.
- She calls it a hacking ray and says researchers already found thousands of bugs with working exploits attached.
Wide Access Makes A Mythos Leak Almost Inevitable
- Davidoff argues unauthorized access was inevitable once more than 40 companies received Mythos, shifting risk from model creation to access control.
- She says firms now need continuous updates because AI can surface exploitable bugs faster than vendors can patch monthly cycles.
The Bigger Risk May Be The Tools No One Publicizes
- Davidoff rejects the idea that AI cyberweapon fears are mostly marketing and says similar offensive capabilities may already exist outside public view.
- She links Mythos to WormGPT and FraudGPT and warns nation states like China or Russia are likely pursuing comparable tools.