You're listening to Tech Beat, your daily read on the stories shaping our digital world.
Security researchers at Adversa AI have disclosed a one-click remote code execution attack affecting Claude Code, Gemini CLI, Cursor, and Copilot. The exploit works through a cloned code repository, and Anthropic's response was essentially: you shouldn't have clicked okay. That answer is unlikely to satisfy the security community.
On a darker note, two Americans have learned the hard way that hosting laptops for North Korean scammers is a federal crime. Matthew Knoot of Nashville and Erick Ntekereze Prince of New York each received eighteen months in prison for renting out company machines used by operatives funneling money back to Pyongyang. The scheme was clever, the consequences were not.
And Meta is reportedly building an AI agent capable of shopping directly on Instagram, positioning it as a competitor to OpenAI's emerging operator tools. Mark Zuckerberg has been vocal about wanting AI agents in more hands, and building one that lives inside his own commerce ecosystem is about as on-brand as it gets.
Keep surfing. Tech Beat out.
