Профиль

This is just to say I have OCRed the text thet was in tne arch:ve and whicb yov were prabaldy seamhing for "breahtast" Fargme ms fheg mem debo:w5 sb ?mccf mmb 5c oc1b

The real English majors won't go for the most common WCW poem for this joke.

See the figure five in gold in UNPARALLELED CLARITY on certified IMAX screens

Real-time footage of Hugging Face's LLM-powered security program.

Dario sees the economic reality in front of him and observes a shift to relying on open-weights models for some workflows. That's a potentially existential risk for his business (as China no doubt intended). The rest is backwards reasoning.

I'm not saying open-weights models are unproblematic. I'm just saying this rationale for their danger doesn't pass a sniff test.

He's spooked. But also, to read this you would think that nobody had ever jailbroken Claude into performing criminal activity. Except...their own reports tell us that's not so. All models can be abused; no models can be secured. Why then should we trust any product we can't examine at all?

Dario's post on open-weights models is intentionally misleading. With one hand, he proffers praise and admiration for the models. With the other, he tsk-tsks about the "danger" of biological attacks, demanding bans on chip sales and penalties for distillation. www.anthropic.com/news/positio...

Finally was able to cohere my thoughts about the OpenAI/Hugging Face debacle. We keep making weapons. We will always want to use them. Can we break the cycle?

I would make a Claudewind wall of shame but it would take all my time just developing the list. The web design would be no CSS, just raw HTML. You get a <ul> and some <li>s and that's it.

Or, I should have said a file read for a Python script. Claude wasn't like "I'm not analyzing a dataset; that sounds SCARY. It tried, then got itself into a situation where the results tripped guardrails.

An important nuance is that the refusal happened during a tool call: a Python invocation that appears to be an encryption routine, complete with XOR (also HuggingFace didn't redact their token but OH WELL). huggingface.co/datasets/hug...

Finally was able to cohere my thoughts about the OpenAI/Hugging Face debacle. We keep making weapons. We will always want to use them. Can we break the cycle?

Update: I was told directly they were in the CVP.

We continue to use our own RSS-Filter project to curate our feed aggregator. We've removed "Startups" from the other-wise excellent TechCrunch security news feed to keep the feed relevant and actionable. Our Newsfeed: news.ifin.network

We now know Hugging Face's AI cybersecurity "program" was running Claude Code Opus 4.8 against some logs. They weren't in the Cyber Verification Program so got a refusal. They moved to GLM 5.2 with the detailed prompt "analyze this dataset." 🤡

My three year-old just started booing Trea after that error and I'm so proud of her.

This site changed my life. Jedi Knight modding was my entry to hacking as a concept.

This is what we're here for! If you see something suspicious and want someone else to check it out, drop us a tip! We take care of each other by sharing what we see.

Following up on a Fediverse tip, we found a new use for fake software download sites: the referral program hustle. Mirror FOSS, get cash. There are almost certainly more of these out there. discourse.ifin.netwo... #ThreatIntel #ThreatIntelligence #IFIN