Gartner predicts 40% of agentic AI projects will be cancelled by 2027, mostly over runaway costs. This week, a startup called Sapiom raised $35 million betting it can fix exactly that problem, backed in part by Anthropic, the same kind of AI lab whose pricing is part of what's driving that cancellation rate in the first place. So is Sapiom solving the problem, or just profiting from it?
Six global cybersecurity agencies just warned hospitals about a risk hiding inside their own AI systems, and it's not hackers they're worried about. AI agents are now drafting care plans, submitting prior authorizations, and coordinating patient care with almost no one signing off on each step. When one of them makes a mistake, the error can quietly cascade before a human ever catches it. So who's actually accountable when it does?