You're tuned in to Tech Beat, let's get into it.
Security researchers have pulled off a striking demonstration of AI-enabled hacking, using Anthropic's Claude to breach OpenAI employee ChatGPT accounts. By chaining two vulnerabilities together, the trio gained access to internal repositories, raising uncomfortable questions about how well the world's most prominent AI labs are protecting their own infrastructure.
That uncomfortable irony lands just as Anthropic's CEO Dario Amodei is pushing what he calls a plan to pace the frontier of AI development. The proposal centers on independent safety evaluators and coordination among labs in democratic nations. It's drawn some industry backing, but Nvidia's Jensen Huang has been notably skeptical, and the core question remains unanswered: who actually enforces any of this?
Meanwhile, China's Salt Typhoon hacking group has shifted its attention toward Latin America, deploying a newly developed backdoor called SparroWocky against governments in Argentina, Peru, Venezuela, and beyond. The tool carries more than thirty commands built for surveillance and data theft. Researchers believe the pivot tracks with renewed American pressure in the region, turning geopolitics into a direct driver of cyber operations.
Stay curious out there. Tech Beat out.
