Welcome to Tech Beat, your daily look at the stories shaping our digital world.
Meta's new AI assistant Muse has hit a rough patch before it even finds its footing. Despite CEO Mark Zuckerberg's bold claims about privacy and security, researchers have uncovered a zero-day vulnerability that hands locally run apps and terminal commands complete control over the agent. Amazon has already responded by blocking Muse from its platform entirely.
That story connects to a broader pattern in AI security. Anthropic has been linked to a growing pile of CVEs through its Project Glasswing vulnerability research initiative, but here's the interesting wrinkle — fewer than half a percent of those disclosed vulnerabilities are actually being exploited in the wild. The gap between finding bugs and attackers acting on them remains, for now, surprisingly wide.
Meanwhile, a quieter but unsettling threat is gaining ground. Hackers are embedding malware instructions directly inside public blockchains, creating communication channels that survive even after traditional servers are taken down. Unrestricted AI tools are reportedly making these attacks easier to execute for less experienced criminals, with incidents surging four hundred forty percent. Blocking blockchain traffic is complicated by the fact that it runs alongside legitimate cryptocurrency services.
Three very different stories, one common thread — the tools we build keep surprising us in ways we didn't plan for. Keep surfing. Tech Beat out.
