Despite being market competitors, OpenAI and Anthropic share concerns about 'open-weight AI models,' which could impact their profitability.
Imagine you are using a “cutting-edge AI assistant” service that costs a significant monthly subscription fee. Then, one day, someone uploads the entire “AI engine file” to the internet, which has performance comparable to your service and is almost free to use. Now, people no longer need to pay monthly fees; they can download that file to their own computers and run the AI themselves. This has effectively sounded an alarm for existing AI companies.
Recently, the two pillars of the artificial intelligence industry, OpenAI and Anthropic, have been raising their voices to policymakers daily. They are highlighting the risks posed by “open-weight AI models”—models where anyone can access the weight data and execute it directly in their own environment. While they are fierce market competitors, they are united in their stance on the impact these models could have on the ecosystem.
Why is this important?
This issue is not just for technologists. Until now, the AI ecosystem has been dominated by the use of “closed models” created by companies like OpenAI or Anthropic through the services they provide (APIs, or Application Programming Interfaces). Source: This Week in AI: Open Source Just Broke Anthropic and OpenAI Business Model
| However, the situation is changing rapidly as open-weight models spread. As users can download models and run them on their own servers or internal corporate infrastructure, they can freely leverage artificial intelligence without going through the APIs of large AI companies. Some analysts point out that this phenomenon could directly pressure the profit margins of these large AI firms. [Source: OpenAI is scared of open-weight models. Should the US be? | TechCrunch](https://techcrunch.com/2026/07/20/openai-is-scared-of-open-weight-models-should-the-us-be/) |
In simple terms
Shall we compare a “closed model” to the kitchen of a famous restaurant? You go to that restaurant, pay money, and order a finished dish (the result). Only the chef knows the secret recipe, and if you want to enjoy that taste, you must go to that restaurant.
On the other hand, an “open-weight model” is like a famous chef revealing their recipe to the public in detail. Now, you can take that recipe and cook the exact same dish in your home kitchen. It is much cheaper because you don’t have to go through the restaurant.
| Large AI companies argue that releasing these models can cause safety issues. Of course, the fact that a powerful competitor to their own model has emerged from a business model perspective is also discussed as one of the industry’s core concerns. [Source: OpenAI is scared of open-weight models. Should the US be? | TechCrunch](https://techcrunch.com/2026/07/20/openai-is-scared-of-open-weight-models-should-the-us-be/) |
Current situation
Anthropic CEO Dario Amodei has warned that releasing powerful AI models without centralized management could lead to uncontrollable risks. Source: Anthropic CEO warned that open-source AI could become “very dangerous” as Western companies quietly switch to Chinese AI models - Tech Startups
However, companies are turning to open-weight models to find more economic alternatives, and reports continue to surface that instances of introducing open-weight models developed overseas are increasing. Source: Anthropic CEO warned that open-source AI could become “very dangerous” as Western companies quietly switch to Chinese AI models - Tech Startups
What happens next?
| Braden Hancock of Snorkel AI predicted that powerful open-source models would pressure the profit margins of existing large companies, ultimately inducing price competition. [Source: OpenAI is scared of open-weight models. Should the US be? | TechCrunch](https://techcrunch.com/2026/07/20/openai-is-scared-of-open-weight-models-should-the-us-be/) |
Ultimately, the more large AI companies try to raise policy barriers, the more likely it is that the use of the technology will spread even further. Watching how policy authorities balance the two values of “corporate safety management” and “public accessibility of technology” will be a key point of interest.
AI’s perspective
AI safety is a very important value. However, we need to look cool-headedly at whether that justification is just confined to the fence of protecting corporate profits. We must not forget that technological innovation starts in an open environment that more people can utilize.
References
-
[OpenAI is scared of open-weight models. Should the US be? TechCrunch](https://techcrunch.com/2026/07/20/openai-is-scared-of-open-weight-models-should-the-us-be/) -
[Anthropic CEO: Open-Source AI is getting dangerous (2023) Hacker News](https://news.ycombinator.com/item?id=48716750) - OpenAI and Anthropic are pulling in different directions - Help Net Security
- OpenAI and Anthropic find common ground: Open-weight AI
- This Week in AI: Open Source Just Broke Anthropic and OpenAI Business Model
-
[Anthropic vs. OpenAI: The Two AI Giants Compared DataCamp](https://www.datacamp.com/blog/anthropic-vs-openai) - Anthropic CEO warned that open-source AI could become “very dangerous” as Western companies quietly switch to Chinese AI models - Tech Startups
-
[OpenAI AI News — Latest Updates, Tracker & Coverage AI Weekly](https://aiweekly.co/ai-news-today/openai-news) -
[It’s not about Anthropic vs. OpenAI anymore TechCrunch](https://techcrunch.com/2026/06/26/its-not-about-anthropic-vs-openai-anymore/)
- The models have low security performance
- Affordable alternative models pressure their revenue models
- They cannot receive government subsidies
- Available only in corporate clouds
- Users can download and execute model files directly
- Usable only in official OpenAI apps
- AI model development costs are too high
- The danger of releasing powerful AI models without centralized management
- Internet speeds need to be faster