How long will the love-hate relationship between Anthropic and OpenClaw last? Last week, Anthropic just officially announced that it had "banned" OpenClaw. This week, it directly "banned" Peter Steinberger, the "father of OpenClaw"... Early this morning, OpenClaw founder Peter Steinberger posted on X: "Yes, everyone, it will be more difficult to ensure that OpenClaw continues to be compatible with the Anthropic model in the future."

In addition, a picture of an email from Anthropic's security team was attached, stating that Peter Steinberger's personal Claude account was suspended due to "suspicious" activity.

Although there is only one short sentence, Peter Steinberger's "anger" and "helplessness" can also be seen.


The following is the full text of Anthropic’s email:

Hello,

We conducted an internal investigation into suspicious signals associated with your account, which resulted in a violation of our usage policies. As a result, we have revoked your access to Claude.

If you wish to appeal, please fill out this form or learn more about the appeals process here.

Sincerely

Anthropic Safeguards Team

But the "ban" did not last long. Just two hours later, Peter Steinberger posted again: "My account has been restored, thank you everyone!"


However, the account has been unblocked, but the aftermath of the incident continues. Since Peter Steinberger posted his post, the popularity has continued to rise, and various conspiracy theories and doubts are flying under the post.

First of all, everyone feels that Anthropic's approach is "too unkind".

At first, some netizens thought that maybe Peter Steinberger was "banned" because he really violated the rules, but after learning about it, he found that this was not the case.

"Is it because you violated their terms of service and used the API in an impermissible way that CPS (Claude Protective Services) was used against you? Of course they will do so. This is their responsibility. According to the terms, such accounts are not allowed to use services through third-party tools."

But Peter Steinberger's answer was, "This is a first-hand call through the API key and using the official Claude CLI."


Some netizens thought that Anthropic might have made a mistake and asked, "Is it the API account that Anthropic banned this time?"

And the answer was "Oh they blocked both." (Both of them were blocked.)


According to Peter Steinberger, since Anthropic released the new rules last week, he has been strictly complying with them, using the official API mode, and pre-depositing funds into his account. He has truly transformed from a "developer who earns money" to a "paying customer with real money", but he is still banned, so that he now "cannot even test."

Of course, some people question, since Peter Steinberger is now working at Anthropic’s competitor OpenAI, why is he still using Anthropic’s products? What is it mainly used for?

Peter Steinberger also felt wronged. Wasn't everything he did to ensure the cross-model compatibility of OpenClaw?

“We’ve been running e2e tests to make sure the changes to Harness won’t cause problems on mainstream models. That’s basically what we use keys for.”


In addition, he also said that there was no conflict.

"You need to separate two things: my job at the OpenClaw Foundation is to make OpenClaw well compatible with all models/providers; and my job at OpenAI is to help them develop their future product strategy."



Some netizens even lamented that if they had not made the choice they did now, things might not have developed to this point. "You could have chosen, but you made the wrong choice..."

Peter Steinberger also fought back without hesitation: "One of them welcomed me, the other sent a legal letter..."

The unspoken inner subtext is probably: Now you should know why I chose OpenAI, right?


As the post became more popular, Thariq Shihipar, an engineer at Anthropic, also "entered the scene". "We're not banning anyone for using OpenClaw," he said. "It's likely that this is triggering another abuse detection classifier that's a little too sensitive. I'll send you a private message and we'll take a look at the situation and help you get unbanned."


I don’t know if it was Thariq Shihipar’s help or if the “ban” was just an own mistake. In short, two hours after the “ban”, Peter Steinberger’s account was restored.

The whole thing has come to an end here, but it is undeniable that the rift between the two parties has grown wider again.

The reason why netizens feel that this ban is not a misunderstanding, but intentional, and looking at the problem from the perspective of Peter Steinberger, is that there is actually a cruel reality. In the era of large models, the "lifeline" of open source software is still in the hands of model manufacturers.

Whether it was the last "ban" or this time, manufacturers can "rectify" it at any time by modifying the terms of service and other methods...


https://x.com/steipete/status/2042615534567457102

https://techcrunch.com/2026/04/10/anthropic-temporarily-banned-openclaws-creator-from-accessing-claude/