Prof G Markets – April 23, 2026
Episode: “The Most Dangerous AI Model Just Leaked”
Hosts: Ed Elson (Vox Media Podcast Network)
Guests: Sherry Davidoff (Founder, LMG Security & co-host, Cyberside Chats), Seth Goldstein (Senior Equity Analyst at Morningstar)
Release Date: April 23, 2026
Episode Overview
This episode provides a sharp, no-nonsense briefing on two urgent market stories: the major security breach of Anthropic’s most powerful AI model, Mythos, and Tesla’s latest earnings report, including their progress with Robotaxi and humanoid robots. The hosts also analyze Iran’s ongoing hostilities and their overlooked impact on global markets. The Prof G Markets team delivers expert context, skeptical inquiry, and brisk market intelligence without jargon.
1. Anthropic’s Mythos AI Model Leak: Security Threats & AI’s New Era
With guest Sherry Davidoff, LMG Security (MM:SS 02:07–14:25)
Key Points & Insights
- Incident Recap:
- Anthropic released Mythos Preview, an extremely advanced AI model, in a controlled, limited rollout to around 40 tech companies.
- Bloomberg reported that unapproved, unidentified users gained access the same day Project Glasswing (the controlled rollout) launched.
- Anthropic is investigating but, as of the recording, found no evidence of malicious use.
- The breach raises severe questions about AI security, access, and the risk of misuse at scale.
Breaking Down the Threat
-
Magnitude of the Risk:
- Sherry Davidoff: “Mythos is completely changing the security risk landscape right now... It’s basically like a hacking ray. You can point it at software, and even if you don’t have the source code, it can still detect vulnerabilities and write code that will exploit it.” (05:55)
- The model can autonomously scan any software for vulnerabilities and generate working exploits without human intervention, giving potentially anyone with access the power to hack at scale.
-
Breach Inevitability & Vetting Concerns:
- With so many organizations involved, unauthorized access was “inevitable.”
- Notable quote: “They say three can keep a secret if two of them are dead.” (Sherry, 04:45)
- Unclear how companies are being vetted, what safeguards are in place, and whether even best-in-class security can prevent further leaks.
-
Industry Implications:
- The old cycle of discovering, patching, and responding to software vulnerabilities can’t keep up. Now there is risk of mass, automated exploitation.
- Small suppliers and businesses could be overwhelmed; rapid, continuous updates and AI-powered defensive tools will become the norm.
-
Role of Government & Regulation:
- Parallels drawn to past government-held arsenal of vulnerabilities (like NSA/CIA “stockpiles”).
- Sherry: “Maybe setting some standards for disclosure. If you have an AI that has certain capabilities, do we need to track that? Do we need to notify anybody?” (10:04)
- Government could/should set disclosure standards for high-risk AI capabilities.
Hype or Reality?
- Skepticism Addressed:
- Ed suggests some of the warnings could be Anthropic’s “fear-mongering marketing tactic” to hype their own technology for investors.
- Sherry rebuts, citing direct experience with “dark web” tools already showing advanced capabilities:
- “I wish I thought it was hype, but I myself have been using Claude code. I’ve been looking at the capabilities of Opus... Tools like Worm GBT and Fraud GPT... were already pretty darn good at finding vulnerabilities and writing exploits.” (11:57)
- The threat is real; security professionals are not overreacting.
Geopolitical & National Security Concerns
- State Actor Interest:
- China, Russia, and Iran are undoubtedly seeking access to such powerful models, raising national security stakes.
- The possibility of insider risk through bribery—e.g., Russian agents once offering a $1 million bribe to a Tesla employee to plant malware—was highlighted as a cautionary parallel. (13:26)
Memorable Quotes
- Sherry Davidoff, 05:55:
“It’s basically like a hacking ray. You can point it at software, and even if you don’t have the source code, it can still detect vulnerabilities and write code that will exploit it.” - Ed Elson, 10:51:
“It would make sense for Anthropic to say, look how crazy and bad this is and get a lot of headlines and generate a lot of heat. But then at the same time, it also seems very dangerous and frankly stupid to just assume that it’s marketing and sort of cover our eyes... Because it could very well be a problem.”
2. Tesla Earnings and Future Bets: Cash Flows, Robotaxi & Robots
With guest Seth Goldstein, Morningstar (MM:SS 17:41–32:23)
Q1 Performance & Market Outlook
-
Earnings Headlines:
- Tesla beat analyst revenue and cash flow expectations, despite Q1 deliveries missing projections.
- Positive free cash flow is crucial given Tesla’s ambitious $20+ billion expenditure on AI and robotics pivot.
-
Highlight Developments:
- Seth: “Tesla was able to launch its Robotaxi in two new markets by April... launching immediately into unsupervised robo taxi [no safety monitor].” (18:31)
- Progress in factory refit: shifting a plant from cars (Model S/X) to producing “Optimus” humanoid robots—on track to start production by year-end.
-
Robotaxi Progress (and Skepticism):
- Robotaxi has underdelivered for years; “always a year away.”
- Latest launches (Dallas, Houston) now feature no safety driver, suggesting meaningful software progress.
- Rollout remains intentionally slow to avoid a repeat of Cruise’s disaster, but there’s “real revenue” and plans to expand to five more cities soon. (21:35)
- Seth: “We’re actually seeing progress. But Tesla wants to go intentionally very slow here... Nobody wants a repeat of the Cruise incident.”
Delivery & Sales Dynamics
-
Auto Business Health:
- Deliveries up 6% YoY, but production outpaced deliveries by 50,000 cars—the biggest gap yet.
- Automotive profit margins remain strong (>19%), signaling core business resilience.
- Growth of high-margin software (Full Self-Driving subscriptions rose another 100,000 in the quarter) and insurance strategies are offsetting some auto market headwinds.
-
Regulatory Tailwinds:
- Pending approval for FSD in Europe may boost sales, as software differentiates Tesla in a competitive EV landscape. (23:44)
Valuation: Is Tesla Still a Car Company?
- Market Valuation Context:
- Trades at 185x forward earnings—far higher than other “Mag 7” stocks (Apple at 30x, for example).
- Market assigns such a rich multiple based on expectations for monumental success with high-margin software (autonomy) and humanoid robots.
- “If the software works... it’s going to be a very high margin free cash flow stream that’s coming.” (Seth, 28:17)
- Morningstar’s fair value estimate is $400—roughly the current trading level—implying neither over- nor under-valued, per their projections.
SpaceX IPO Ripple Effects
- Potential Impact:
- SpaceX may soon go public at a $2 trillion valuation.
- Question: Will it siphon off the “Elon Musk premium” from Tesla if investors split their capital?
- Seth argues excitement and index inclusion will allow many to hold both, especially as the two companies could become increasingly synergistic (e.g., Tesla providing robots and solar/batteries for SpaceX infrastructure).
Memorable Quotes
- Seth Goldstein, 21:35:
“The Robotaxi has been a year away for most of the past decade and so I understand the reason for the skepticism. But we’re seeing tangible progress now... They want to be confident in the technology, confident that the software can handle all of the edge cases...” - Ed Elson, 26:07:
“Can they come back, especially in Europe, from this massive slump that we did see in sales where everyone basically just turned against Tesla as a company essentially because of Elon?” - Seth Goldstein, 28:17:
“If the software works, then it’s going to be a very high margin free cash flow stream that's coming.”
3. Iran Ceasefire Extension: Market Complacency & “News Fatigue”
(MM:SS 32:23–36:59)
Breakdown
-
Ceasefire Realities:
- The so-called ceasefire between the US and Iran has been repeatedly violated by both sides.
- Both US and Iranian forces have attacked ships and continued military actions despite public claims to the contrary.
-
Investor Complacency:
- Markets have rebounded as if the situation is normalizing, but Ed Elson argues nothing fundamental has changed—volatility remains, oil and gas prices are still high.
- “We are entering the next chapter of the story... news fatigue, where the headlines become overwhelming, the news becomes confusing, and eventually we get bored of it and decide to stop caring altogether.” (34:44)
-
Core Point:
- The only thing that’s changed is our “tolerance” or “normalization” of a continued state of conflict, a dangerous precedent which could blindside markets if the situation worsens.
-
Ed’s crucial question:
- “What exactly has changed? ... The way I see it, nothing has changed. We’re still at war with Iran.” (33:17, 34:18)
Timestamps for Important Segments
- [02:07] – [14:25]: Mythos AI model leak & cybersecurity stakes (Guest: Sherry Davidoff)
- [17:41] – [32:23]: Tesla Q1 earnings, Robotaxi/robotics, valuation & SpaceX IPO effects (Guest: Seth Goldstein)
- [32:23] – [36:59]: Iran “ceasefire,” market psychology, and the danger of news fatigue
Tone and Commentary
- Ed Elson maintains a brisk, direct, slightly sardonic style—blending market skepticism with humor (e.g., “Money market matter. If money is evil, then that building is hell. The show goes on.” – 02:07).
- Expert guests cut through hype and PR spin, foregrounding real risks and operational realities.
In Summary
This episode delivers a cogent, high-alert analysis of the collision between AI innovation and security risk, the transformative ambitions (and volatility) of Tesla, and markets’ tendency to look away from ongoing geopolitical peril. The message: the ground is shifting beneath the surface of tech and markets, even if investor behavior lulls us into complacency.
Notable Quote to End:
Ed Elson (34:44):
“We have decided for our own sanity that things are improving or at least that things don’t really matter anymore... I think the economy is about to tell us a hard truth. They haven’t.”
