Experts cited by the Financial Times report that ChatGPT, Gemini, and other Western AI tools are actively accelerating Iran's ability to develop malware and execute phishing campaigns, despite platform safeguards
For AI PMs in regulated environments, this is a direct signal that model misuse is an enterprise reputational and compliance risk — expect tighter usage-policy scrutiny and pressure from security and legal teams to audit how your AI tooling could be weaponised or implicated in adversarial contexts
Reuters Breakingviews reports that Anthropic calculates "run-rate revenue" by annualising the last 28 days of consumption-based sales (×13) and adding annualised subscription revenue (×12), a non-standard methodology flagged by an insider
AI PMs evaluating Anthropic as a strategic vendor should probe these financial metrics carefully — non-standard reporting can obscure true growth trajectories and matters when justifying multi-year platform commitments to finance and procurement teams
Microsoft is ending flat-rate Copilot subscriptions in favour of token-usage pricing, prompting widespread developer backlash over dramatically higher projected costs, particularly for heavy users
AI PMs championing developer tooling must urgently model consumption patterns before this change hits budgets; token-based billing also signals an industry-wide shift that will require new cost-governance frameworks for any AI coding assistant rollout
Blogger David Wilson, highlighted by Simon Willison, documents spinning up 16+ unplanned AI-assisted projects, concluding that AI makes it trivially easy to start things but does not reliably solve the original problem — and Willison finds the pattern deeply relatable
For AI PMs, this is a cautionary signal about measuring AI productivity by output volume rather than outcomes; in regulated environments, unchecked AI-driven project proliferation also creates shadow IT and compliance exposure
The Reuters Breakingviews methodology breakdown (via Willison) is a short but illuminating read on how AI companies construct growth narratives using non-GAAP-style metrics that can mislead investors and enterprise buyers alike
PMs presenting AI business cases internally should anticipate similar scrutiny from CFOs and audit teams when citing vendor revenue figures as a proxy for stability
Simon Willison's reflection on David Wilson's project-sprawl post digs into a genuine tension in AI tooling: the technology is optimised for starting things, not finishing them, and may reward exploration over execution in ways that are hard to justify in a corporate ROI context
Worth sharing with product and engineering leads as a grounding read before your next AI tooling budget review