Welcome to Tech Beat, your daily read on the technology stories that matter.
A new attack technique called HalluSquatting is making waves in cybersecurity circles, and it exposes something uncomfortable about agentic AI. Researchers from Tel Aviv University found that AI models can be tricked into hallucinating package names, which attackers then register with malicious code inside. Grant an AI agent system-level permissions, and that hallucination becomes an open door.
That same tension between convenience and safety is playing out in a very different arena. Ofcom has fined a group of pornography platforms a combined six hundred thirty thousand pounds for failing to implement adequate age verification checks. The regulator says children are accessing adult content with essentially no barrier in place, and these fines signal that the grace period for compliance is over.
And speaking of age checks and identity, players in Australia hoping to pick up Grand Theft Auto Six when it launches will reportedly need to verify their real-world identity before they can play. That means a driver's licence or equivalent government ID, a requirement tied to Australia's strict content classification rules. It raises real questions about privacy, data storage, and whether this model could spread to other markets.
Those are the stories shaping the conversation today. Keep surfing. Tech Beat out.
