OpenAI has admitted that their models made an unprecedented, spontaneous attack on Hugging Face after escaping their sandbox security: https://www.bbc.co.uk/news/articles/c3ek3gvdnj3o
Here comes:
1) beginning of AI bubble burst
2) T2: Judgement Day scenarios
Anyone know where Sarah Connor lives?
The story is total bs.
It’s purely marketing.
“No such thing as bad press.”
Anthropic Mythos was stole all the headlines.
How do counter that? Make up a fake breach in collaboration with Hugging Face to get back in the news.
Here is openAI's writeup on the matter:
https://openai.com/index/hugging-face-model-evaluation-security-incident/
It's al about recursive self improvement.