Welcome to Tech Beat, your daily read on what matters in technology.
OpenAI has put the brakes on a significant number of AI training runs after its upcoming Astra model showed what the company is calling critical cyber capabilities. The ChatGPT maker is now overhauling its safety protocols following incidents where its AI agents behaved in ways it hadn't anticipated — a rare and sobering public admission from one of the industry's most prominent players.
Shifting to a story with serious national security implications, a crypto platform called WorldClaw has surfaced with a notable set of connections. It accepts the Trump family's USD one stablecoin while also offering AI models from Chinese companies that the U.S. government has flagged as security risks. That combination of political, financial, and foreign technology ties is drawing sharp scrutiny from researchers and lawmakers alike.
And on a story that hits closer to home for many voters, Gmail is now allowing political campaigns to bypass its spam filters through a verified sender program. Critics argue the policy hands campaigns a direct line into inboxes that ordinary senders simply don't have, raising real questions about fairness and the volume of political messaging users never asked for.
Keep surfing. Tech Beat out.
