Anthropic and OpenAI May Tighten API Access, Putting Enterprise Clients Under Dual Pressure
Nashnova编辑部
Anthropic and OpenAI are pushing into vertical applications, competing directly with the developers who rely on their APIs — design tool Canva is the latest caught in the crossfire. This means → companies building on frontier models now face a double risk: their supplier becoming a rival, and their access being curtailed.
What does "your supplier becomes your competitor" actually look like?
Anthropic and OpenAI are no longer just selling model access — they are building their own vertical apps, competing head-on with API customers.
Design tool Canva is the latest example: its core features now overlap directly with Anthropic's own applications.
This means → a developer's most critical technology vendor can become its biggest competitor overnight — while the product still runs on that vendor's model.
Why is API access being tightened — safety and anti-distillation?
The Trump administration has pressured both companies to adopt staged releases for new models, requiring customer-by-customer approval before granting access.
Even after public release, both firms may deliberately degrade model performance on specific tasks — Anthropic's Fable model has already been weakened on cybersecurity-related capabilities.
In plain terms = outside customers get a nerfed version, while the companies themselves run the full model for their own apps — an inherently uneven playing field.
What is "model distillation," and why does it add to the pressure?
Model distillation — training a smaller model on a larger model's outputs so it inherits some of its capabilities — is a common industry practice.
Anthropic warned in a risk report that distilled models may inherit dangerous capabilities without the original model's safety guardrails.
Yet a former Anthropic researcher told The Information that fully preventing distillation is "basically impossible" — this reflects a natural ceiling on technical lockdowns.
Is commercial interest the real motive for restricting access?
Some investors have already warned developers: Anthropic may reserve its most advanced technology for its own competing apps rather than offering it through the API.
Anthropic has moved into AI-driven drug discovery and other high-margin verticals.
This means → if Anthropic follows through, it would face a major antitrust investigation — but the potential returns from vertical apps may far exceed API revenue.
How big is Anthropic's API business right now?
Bloomberg reports Anthropic's Q2 revenue at roughly $11.5 billion, up approximately 14× year-on-year (from $787 million a year earlier; Q1 was $4.73 billion).
The company has reached profitability on an adjusted operating-profit basis.
In plain terms = the API business is in a hyper-growth phase and walking away would be enormously costly — but if owned apps can generate even more, the calculus shifts.
How are developers responding?
Legal-AI firm Harvey and code editor Cursor have both begun training proprietary models to reduce dependence on Anthropic and OpenAI.
The Information expects more developers to follow the same path.
This reflects a deeper signal: when your core infrastructure provider can become a competitor at any moment, building in-house capability stops being a slogan and becomes a survival strategy.
Content is for reference only, not financial advice.