Bloomberg News Now
THE SO WHAT
AI safety moving onto the UN General Assembly agenda means frontier model governance is now a diplomatic topic, not just a regulatory one. Multinationals should expect cross-border AI rules to harden — and design architectures that can respect divergent national constraints.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIWindows 11 Copilot app is getting its own browser, but not everyone's happy
Bundling a Copilot-specific browser is Microsoft treating the assistant as a primary navigation layer, not a sidebar feature. If you build on Windows, assume more user journeys will start inside Copilot’s surface rather than your app or the default browser.
Applied AI‘You cannot hide from AI’: Should we fear AI-led performance reviews?
AI-led performance reviews move the power center from managers’ judgment to data exhaust and model design. HR and ops leaders need to decide now who owns the training data, appeals process, and bias audits before these systems quietly become the default.
Applied AIThe case for purpose-built generative AI in fraud prevention
Fraud teams are learning that generic LLMs tuned for chat are a poor fit for adversarial, high-stakes detection. If you run risk or compliance, push vendors on domain-specific models, data access, and feedback loops — not just ‘we added gen AI’ marketing.
Applied AIAnthropic reported disrupting a Beijing-aligned operation to monitor religious leaders and diaspora communities, including a DC-based advocacy group for Tibet
State-aligned actors using AI to monitor diaspora communities is a reminder that model providers now sit on the front line of geopolitical surveillance. Any organization deploying advanced AI should assume their tools can be dual-used — and invest in abuse detection and escalation paths, not just uptime.