Welcome Back. Anthropic’s public release of Claude Fable 5, an updated version of the Mythos artificial-intelligence model that spooked cybersecurity experts, is drawing the ire of AI developers over its blunt safety barriers.  

In private testing, the original Mythos was able to identify thousands of vulnerabilities in digital systems at machine speed, much faster than security teams could ever patch them. Anthropic said the model was too dangerous to release widely.

Now, when a user touches on sensitive topics like bioweapons and cybersecurity, Fable pops up a notification and redirects the conversation to an earlier, less capable model.

Fable also degraded the quality of its responses about high-end AI development to be less useful for developers looking to build AI tools that might not have the same safeguards. For these responses, there was no pop-up notification, however. The company cited national security and its own terms of service as reasons for the invisible restrictions.

“This is one of the first times that an AI company has rolled out a guardrail and there has been uniform disdain,” Sayash Kapoor, an AI researcher at Princeton University, told WSJ’s Sam Schechner.

Read the full story here. 

Also today:

• Novo Nordisk patient data breached
• FIFA World Cup draws hackers
• EU partners with Brazil on cyber

The European Union has agreed to a digital partnership with Brazil, part of a broader effort to reduce the bloc’s reliance on U.S. technology, Henna Virkkunen, the European Commission ‌tech chief, said on Thursday. The partnership includes cooperation on issues relating to data, connectivity and cybersecurity, as well as the protection of minors. (Reuters)

Siemens is notifying customers that legitimate files for its building-management system are being flagged as malware by multiple cybersecurity tools. The company’s Desigo CC system integrates lighting, security, fire safety, power and other building subsystems into a single platform for centralized monitoring and control. Siemens said it’s working with cybersecurity vendors to address the issue. (SecurityWeek)

The WSJ Pro Cybersecurity team is Deputy Bureau Chief Kim S. Nash and reporters Angus Loten and James Rundle. Follow us on X @WSJCyber. Reach the team by replying to any newsletter you receive or by emailing Kim at kim.nash@wsj.com.