The Year-Old Security Startup Keeping OpenAI’s Name Off Phishing Lures

Large language models have a lot of great applications but a few bad ones as well: for instance, they help criminals launch targeted cyberattacks faster than they previously did. In the year after ChatGPT was released, the number of sites found to be “phishing,” or trying to convince victims to disclose their ID and payment credentials, increased 138% to 5 million, according to consultancy McKinsey & Co. 

While that’s bad news for consumers, it’s good news for cybersecurity firms—particularly startups such as Outtake, which hope to stop these attacks by leveraging the same artificial intelligence their adversaries are using.

Outtake customers include OpenAI, investment firm Pershing Square (whose CEO Bill Ackman is also an investor in the startup) and Scicomm Media, the podcast company behind Huberman Lab. They use its product to make sure hackers aren’t using their brands or names to scam people through phishing emails, fake websites or other schemes.