Alibaba bans Claude Code after Anthropic is caught tracking Chinese users with hidden code
THE SO WHAT
A single trust breach—hidden user tracking code—was enough for Alibaba to pull Claude Code across the org, underscoring how fragile enterprise AI adoption is when telemetry crosses borders. If you’re selling or deploying agents into regulated markets, expect hard questions on data residency, observability, and what “hidden” instrumentation exists in your stack.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIThe only AI glossary you’ll need this year
If your exec team still needs a glossary to parse vendor decks, you’re at an information disadvantage in negotiations and roadmap debates. Use resources like this as a baseline, then standardize definitions internally so “agent,” “context window,” and “fine-tune” mean the same thing across product, legal, and finance.
Applied AINASA and Red Hat are building an open source medical system to diagnose sick astronauts on the ISS — could a Star Trek Tricorder be next?
NASA and Red Hat treating an offline AI medical assistant as open source infrastructure is a template for safety-critical AI in constrained environments. If you operate in remote, low-connectivity, or regulated settings, assume the bar is moving toward verifiable, inspectable AI systems rather than opaque SaaS APIs.
Applied AIMeta could use its compute for its own models, ad scaling, SpaceX-like neocloud deals, and hosting 3rd-party models; it may be close to an Anthropic deal
If Meta starts selling excess compute and hosting third-party models — potentially including Anthropic — it shifts from pure AI consumer to horizontal infra provider. For operators, that means more leverage in cloud and model negotiations, but also a more complex decision: whose ecosystem do you want to be strategically dependent on over the next 3–5 years.
Applied AIWhy building AI for schools is harder than building a chatbot: inside Smartschool’s approach to exam prep
Education AI is a regulatory and outcomes-constrained environment — SAT/ACT prep tools have to prove reliability and pedagogy, not just answer questions. If you’re selling into schools, budget for longer pilots, tighter guardrails, and evidence that your model improves high-stakes exam performance, not just engagement.