The Most Dangerous AI Model Just Leaked artwork

The Most Dangerous AI Model Just Leaked

Prof G Markets

April 23, 2026

Ed Elson is joined by Sherri Davidoff to analyze the security breach of Anthropic’s Mythos, what it means for AI safety and cybersecurity, and how both the company and regulators should respond. Then, Seth Goldstein helps unpack Tesla’s first-quarter earnings.
Speakers: Seth Goldstein, Ed Elson, Sherri Davidoff
**SPEAKER_1** (0:00)
Support for today's show comes from Attio. Attio is the AICRM that gets you from sign up to Insights in seconds. You can connect your emails, calls, product data and more. And Attio can instantly build your pipeline with enriched data and complete context. Whether you're running product led growth or enterprise sales, Attio adapts to your business. Then, ask Attio to plan your next move. Spot deals at risk before they go sideways. Know what to say before every call, and never miss a follow up. Attio will search, update and create across your data instantly so your customer context becomes more powerful than ever.
Ask more from your CRM. Ask Attio and go to attio.com/profg, and you'll get 15 percent off your first year. That's attio.com/profg.

**Seth Goldstein** (1:14)
Or at adobie.com/do that with Acrobat.

**SPEAKER_3** (1:19)
Support for the show comes from Amazon.
There are the things you can plan for. A first birthday party, a movie marathon, a renter-friendly bathroom reno. And then there are the things you can never plan for. A surprise rainstorm, a Blu-ray player calling it quits, stick-on tiles that looked way better on the package. For all things planned and unplanned, Amazon has you covered. You'll find low prices on everyday essentials and last-minute lifesavers. Shop Amazon and save on essentials.
Save the everyday.

**Ed Elson** (1:56)
Today's number, 62 That's the percentage of influencers who report feeling burned out, according to a new content creator mental health study. Today's other number is zero. That's the percentage of Americans who care.
If money is evil, then that building is hell.
Welcome to Prof G Markets. I'm Ed Elson. It is April 23rd. Let's check in on yesterday's market vitals.
The S&P 500 and NASDAQ hit record closing highs on the ceasefire extension. The Dow also rose. Meanwhile, the price of Brent crude rose back above $100 per barrel after Iran seized container ships. The dollar and treasury yields climbed. And finally, cannabis stocks soared on news that the DOJ is moving to reclassify marijuana as a less restricted substance.
Okay. What else is happening? Anthropic built an AI model so dangerous, they refused to release it to the public. We discussed that last week. However, we just learned that unauthorized users got a hold of it anyway. Bloomberg reported that a small group of unknown, unapproved users have had access to the model Mythos for two weeks. They broke into the model the same day Anthropic launched Project Glasswing, a controlled rollout giving a select group of companies exclusive access to the model. Anthropic says it is investigating the security breach from these unidentified individuals but has found no evidence of malicious use so far. The question is, if they can access Mythos without permission, then who else can and how bad could it get? Here to help us answer those questions, we're speaking with Sherri Davidoff, founder of LMG Security and co-host of the podcast Cyberside Chats. Sherri, thanks for joining us on the show here. We discussed Anthropic's Mythos model the other week. It's extremely powerful. How powerful? I don't know, but I know that it was so powerful that they refused to release it to the public. I also know it was so powerful that we saw meetings at the White House where they convened among some of the most powerful people in the nation to figure out what to do about this thing. Now, we learned that someone accessed it, I guess, and it was unauthorized. I can't tell how much of this is real and how important it really is. That's why I have you. What do you make of this?

**Sherri Davidoff** (4:35)
Well, I mean, this is something we've been expecting from day one. They say three can keep a secret if two of them are dead. Mythos Preview has been released to over 40 tech companies, and Lord knows how many people at those tech companies. So from day one, I'm sure there was unauthorized access. The question is, who is accessing it when they shouldn't be? Is it just people at those companies? Is it someone who has wormed their way in from the outside and has actually hacked in? But honestly, the part that I didn't expect was that we would actually hear about it. So bravo to Anthropic for investigating and for whoever it is that figured this out and has started to publicize it.

**Ed Elson** (5:14)
Well, it wasn't them. It was Bloomberg. It was the reporters over at Bloomberg who figured this out.
I mean, I guess if we could just go back a little bit, from my understanding, the problem with Mythos is that it can basically hack any cybersecurity infrastructure anywhere in the world. And you are an expert in cybersecurity. Could you just take us through what is the danger of Mythos exactly and how powerful is it? What are the capabilities?

28 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000763219146

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000763219146