You may have seen a lot of discussion online about Anthropic, Claude, the U.S. government and OpenAI — and not all of it is clear or accurate. Here’s a concise summary of what’s been happening, why it’s significant, and a chance to hear your view. 📌 What triggered the dispute Earlier this year, the U.S. Department of Defense (Pentagon) demanded that Anthropic grant unrestricted access to its Claude AI for military use, including lawful purposes that could include domestic surveillance or autonomous weapons deployment. Anthropic’s CEO, Dario Amodei, refused, citing ethical guardrails the company has built into its AI — especially prohibitions on mass surveillance and fully autonomous weapons without human oversight. When Anthropic stood firm, the Pentagon threatened to cancel a significant contract (≈ $200 million) and label the company a “supply chain risk.” Shortly thereafter, President Trump directed all U.S. federal agencies to stop using Anthropic’s AI technology, citing national security concerns. Anthropic has publicly stated it plans to challenge the “supply chain risk” designation in court and maintain its ethical stance. CEO Amodei characterised his decision not to relent as compatible with defending democratic values — even if it means falling out with government authorities. Meanwhile, OpenAI struck a separate deal with the Pentagon to supply AI for classified military networks, which some commentators believe reflects a different approach to the same set of ethical red lines. 🤔 Why this matters for anyone using AI tools This isn’t just about military contracts — it highlights key tensions in our industry: 1. AI ethics vs. use case pressureCompanies are being pushed to bend their internal safety guardrails in the face of external expectations about how their AI should be used. The Anthropic standoff shows how far a company might go to stick to its principles — and the real consequences when it does. 2. Competition and positioningAnthropic is also promoting easier switching to Claude, even offering tools to import chat histories — part of a broader strategy to attract users amid the controversy.

Mark, the question that was asked in todays session was "Some clients are very sensitive about accountants using AI What ways can we answer their concerns?". You had said you would elaborate on this in the community. Thank you.

Hello all, right after I joined last Fall I had an accident and am only now in the final stages of recovery. I have TONS to catch up on, but I was inspired by the possibilities in Mark's live session today. I'm so grateful for this guidance. Hoping to be active again soon.

So now that I just joined the academy, not an hour later I get this email: Here’s How the AI Crash Happens. Read the full article. @Mark Wickersham do you also follow these types of reports on AI?

If you’ve tried GPT-5 and it’s not quite clicking for you, there’s good news… you can still access the older models. OpenAI has kept them available for Plus, Pro, Team, and Enterprise users—you just have to switch them on. How to do it: 1. Go to your ChatGPT settings (on web) 2. Find General > Show legacy models 3. Toggle it on You’ll then see the older models available again in your model selector. 📸 (I’ll add a screenshot below so you know exactly what to look for) Have you tried GPT-5 yet? Do you prefer it… or are you switching back to one of the older models?