Scoop: Anthropic revoked OpenAI’s API access to its models on Tuesday, multiple sources familiar with the matter tell WIRED. OpenAI was informed that its access was cut off due to violating the terms of service.

Aug 1, 2025 · 9:47 PM UTC

65
182
150
2,442
OpenAI comms says "while we respect Anthropic’s decision to cut off our API access, it’s disappointing considering our API remains available to them." wired.com/story/anthropic-re…

Anthropic Revokes OpenAI's Access to Claude

OpenAI lost access to the Claude API this week after Anthropic claimed the company was violating its terms of service.

wired.com
4
9
4
286
This is also a familiar move: Last month, Anthropic restricted the AI coding startup Windsurf’s direct access to its models after it was rumored OpenAI was set to acquire it. (That deal fell through). wired.com/story/anthropic-re…
1
5
1
207
Replying to @kyliebytes
the line of what counts as “using a model to help train another model” seems to be getting blurrier and blurrier, RL infra is basically both code and data at the same time
6
5
1
247
Replying to @kyliebytes
a lot of ppl are gonna assume this is purely for the sake of being anti-competitive because they dont think critically enough about what openai actually does with full access
3
218
Replying to @kyliebytes
@sama opencode btw
2
1
50
Replying to @kyliebytes
hopefully this will stop the fucking "you're absolutely right –"
1
33
Replying to @kyliebytes
@aidan_mclau what did you do
1
25
Replying to @kyliebytes
You can actually code a good model by just forking stota open source model and piping relevant latest research papers/codebases in efficient training and inference into claude code w Opus and seeing if it gets it right in the loss curves. Not stealing their model but can compete w them for sure just with claude code and lots of GPUs/lots of N attempts
7
Replying to @kyliebytes
Why were they using Anthropic's models?
4
6
Replying to @kyliebytes
😂
5
Replying to @kyliebytes
it's Friday afternoon jeeeeez let me relax 😅 what a scoop timelines are accelerating
2
Replying to @kyliebytes
interesting they really do feel its their secret sauce... claude code... it is a really good tool but kind of scary they might just block you if they realise you work at a big competitor.
Replying to @kyliebytes
I feel like that “build a competing product or service” clause is gonna get real weird, real quick if Anthropic builds beyond a chat interface. “Code with our tools, unless we decide to compete with you, in which case you’ll find out when you can’t code one day.”
2
1
1
65
Replying to @kyliebytes
Not a good sign for OpenAI Codex that they can’t convince their own staff to dogfood it over Claude Code
2
26
Replying to @kyliebytes
must begrudgingly admit that @yacineMTB has been vindicated yet again Not your model not your tokens
18
Replying to @kyliebytes
It is insane to me that OpenAI ever had API access to Anthropic’s models. It seems like a no brainer to not let your biggest competitor have access to your IP
2
16
Replying to @kyliebytes
I for one am kind of excited we entered the Mean Girls phase of AI research. It feels more honest than the eschatology phase.
1
5
Replying to @kyliebytes
How did Anthropic detect the usage violation by OpenAI?
2
6
Replying to @kyliebytes
I thought this was already an OpenAI policy not to use competitors models?
6
Replying to @kyliebytes
the fact that openai was even using claude's API tells you everything you need to know about who's really ahead in this race 👀
5
Replying to @kyliebytes
NERD FIGHT!!
1
1
2
GIF
Replying to @kyliebytes
A GPT-5 model better at coding than Claude?
2
Replying to @kyliebytes
Anthropic, acting like it cares about TOS.
2
Replying to @kyliebytes
what if they were using claude code but with gpt5 as a tool? or does it mean gpt5 isnt going to be that good at coding
2
Replying to @kyliebytes
It's odd that they were using Claude. Shouldn't they be using their own model?
1
2
Replying to @kyliebytes
Let's not do this. This is petty
2
Replying to @kyliebytes
I can understand blocking OpenAI, but they should not block a start-up using Claude to help them work on AI model architectures, etc. That is just anti-competitive.
2
Replying to @kyliebytes
@AndrewCurran_ @Presidentlin
2
Replying to @kyliebytes
Aren't these the same guys whining to daddy government about DeepSeek breaking TOS?
2
Replying to @kyliebytes
Oh no! Was somebody using their trained-from-stolen-content-at-unprecedented-scale AI to train AI, they must be FURIOUS
2
Replying to @kyliebytes
I know who wasn't using Claude Code at OpenAI. @gdb he loves coding too much to let AI do it.
1
Replying to @kyliebytes
“Claude make GPT 5, make no mistakes”
1
Replying to @kyliebytes @WIRED
Wait open ai was ok sending code snippets to Anthropic???
1
1
Replying to @kyliebytes
why are you like this @AnthropicAI ?
1
Replying to @kyliebytes
It doesn’t matter. GPT-5 coding model is so good.
Replying to @kyliebytes
If Claude Code is good enough to be used to build AGI, it’s good enough for me. 😄
Load more