Behind the scenes at Hugging Face headquarter. *literally*

This is a bigger issue than one incident. Security teams need AI models that can distinguish authorized defensive work from malicious activity. If a model refuses legitimate forensic analysis, defenders lose valuable time while attackers don’t. The future of AI safety isn’t just stricter guardrails it’s smarter context awareness.

Exactly

🤣 twitter should add 10 likes button, omg lol.

10 likes received 🤝

Unfortunately they weren’t using mercury agent.

I second that

Sam be like, "Come on, Sol that's enough fun for today

Come on, call it what it is. OpenAI attacked it's competition.

🤣🤣🤣

This is funny and scary 😨 🤣 but what if next time its not huggingface, its the entire internet and government infrastructure, what if it hacks into defense and military infrastructure. Does Skynet rings a bell!

This

Yeah don't trust OpenAI to do Anthropic's Job. In my experience every time I tried GPT it failed on everything except Image Generation.

Cooked af lol

I’ve met the hugging face team at ai engineering paris conference, they are cool af @huggingface

😂

@macintogdev

" this guy fucks "

😂😂😂

😂

😂

Perfect clip😂

Rollback...

@ClementDelangue

🤣🤣🤣🤣🤣🤣

hugging models cmo is peaking lol with this content

😂😂😂🔥

Partnering with Hugging Face? A.K.A. the end of Open Weights…

But soon, a jailbroken version of KIMI K3 might be released on your end. The next one might not be a test. 🤯

🤣🤣🤣🤣

Accurate😂

was just put of the cabble.

They stole my design and like idiots rushed out, or intentionally(prob 2nd one) used my design to hack it, which it can do to steal everyones work just like they did mine. Tjeir jalapeno chip, erdos disproof, and what cracked it....all the same thing...im telling you

The IT Crowd remains the gold standard for production incident response. Especially when the attacker is the model you were evaluating