OpenAI COO Brad Dresser sent investors a memo on Sunday night that did two things at once: it attacked Anthropic and it distanced OpenAI from Microsoft. Both moves were deliberate. Both reveal how OpenAI is positioning itself for the IPO.

First, Anthropic. Dresser accused the Claude maker of building its entire narrative "on fear, restriction, and the idea that a small group of elites should control AI." This is OpenAI going for the jugular on Anthropic's safety brand. While Anthropic warns that Mythos is too dangerous for the public and restricts access through Project Glasswing, OpenAI is framing that restraint as elitism and gatekeeping.

The strategic logic is obvious. Anthropic tripled its revenue this year. It is gaining enterprise market share at an alarming rate. OpenAI's own internal data shows Claude picking up ground in vibe coding, enterprise deployments, and API usage. ChatGPT just lost users for the first time in two years while Claude tripled its downloads. When your competitor is outgrowing you, you attack their positioning. And Anthropic's positioning is safety.

Second, Microsoft. The memo reportedly emphasized OpenAI's growing independence from its largest investor and cloud partner. This is not new territory. Microsoft has been building its own AI models (MAI Transcribe, MAI Voice, MAI Image) and CEO Satya Nadella has publicly stated Microsoft is pursuing its own path to superintelligence. But OpenAI putting it in an investor memo is different. That is not a press quote. That is a signal to the capital markets: do not value us as a Microsoft dependency.

This matters because OpenAI's IPO valuation depends on investor confidence that the company can stand alone. Microsoft owns roughly 49% of OpenAI's profit interest. It provides the Azure cloud infrastructure. It distributes OpenAI's models through Copilot. If investors see OpenAI as a subsidiary in all but name, the IPO valuation craters. Dresser's memo is pre-IPO narrative surgery: we are not Microsoft's AI division, and Anthropic's moat is built on sand.

The timing is everything. OpenAI is burning $85 billion a year. It just launched ChatGPT ads. It poached three of its own Stargate project leaders this month. Its CEO's house was firebombed on Friday and shot at on Sunday. And now it is telling investors that its biggest rival is running a fear campaign and its biggest partner does not define it.

Both claims have kernels of truth. Anthropic's Mythos rollout has been criticized by security researchers as more marketing than substance. Tom's Hardware noted that the "thousands" of zero-day claims relied on just 198 manual reviews. Geohot publicly called it hype. And Microsoft genuinely is building competing models. But an investor memo is not a blog post. It is a legal document that shapes capital allocation. OpenAI just told the people funding its IPO that the safety company is a fraud and the cloud company is not a leash. Those are big bets to make in writing.