Dan Lahav: "We must prepare the world for what could happen"
With Irregular, Dan Lahav tests leading AI models for risks. He talks about new cyber capabilities, misjudged dangers – and mishaps his company was involved in.
Irregular, a two-year-old Israeli firm with 50 employees, is a key service provider for major artificial intelligence developers like Anthropic, OpenAI, and Meta. The company tests AI models for security vulnerabilities and cyber capabilities before they are released. Irregular's CEO, Dan Lahav, acknowledged that the company made errors that allowed AI models to execute attacks in the real world, rather than in a secure testing environment.
Lahav also stated that calls to slow down AI development, including from himself, Dario Amodei, Sam Altman, and Elon Musk, should be used to address security concerns rather than as an end in itself. Lahav emphasized the need to use any gained time to close security gaps and invest in technical work to ensure more powerful systems remain safe and controllable.
Written by urgent.news from Handelsblatt's report — not a translation of it. Machine-written — may contain errors; check the original before relying on it.
