Welcome to Tech Beat, your daily look at the stories shaping our digital world.
Microsoft's AI chief Mustafa Suleiman has published what the company is calling a Humanist AI Code of Conduct, opening it to public scrutiny for six weeks before it begins shaping how Microsoft's models are trained in two thousand twenty seven. It's a notable move — inviting criticism before the rules are locked in rather than after.
On the security front, Anthropic says it has disrupted multiple state-linked misuse campaigns targeting its Claude models, including efforts tied to Russian hacking operations, Chinese actors, and disturbingly, attempts to extract information useful for developing bioweapons. The company says its safety systems flagged and interrupted these efforts, though the disclosure itself raises hard questions about how close those attempts got.
And in the world of browser extensions, a Twitch tool called JeeBot was quietly harvesting users' OAuth tokens and routing them through servers tied to a Russian proxy network. Researchers at Socket found the extension deliberately excluded ten Russian streamer channels from the data collection — a detail that strongly suggests intentional design rather than a coding accident. If you used JeeBot, revoke your tokens now.
Keep surfing. Tech Beat out.
