Welcome to Tech Beat, your daily look at the stories shaping our digital world.
Anthropic has acknowledged that Claude, its widely used AI assistant, genuinely degraded in quality over the past several weeks. A confluence of system changes and software bugs created real regressions in output quality — meaning users who complained weren't being overly critical. The company has published a postmortem and says fixes are underway.
Meanwhile, the White House is raising alarms about artificial intelligence theft at scale. A memo from science and technology adviser Michael Kratsios accuses firms, primarily in China, of systematically distilling American AI models — essentially using the outputs of US-built systems to train competing ones without authorization. It signals that AI intellectual property is becoming a sharper front in the broader tech rivalry between Washington and Beijing.
And in a story that blurs the line between national security and financial markets, a US Army Special Forces master sergeant has been arrested for allegedly using classified intelligence to place profitable bets on a prediction market. The soldier reportedly wagered on the capture of Venezuelan president Nicolás Maduro using information he had access to through his military role — marking what appears to be the first arrest in the United States for insider trading on a prediction platform.
Keep surfing. Tech Beat out.
