Top all-time · #12789
Mithril Security Blog
Company blog · blog.mithrilsecurity.iosecurityai
Mithril Security Blog is the company blog at blog.mithrilsecurity.io, writing about security and ai. Since 2022, 6 different Hacker News readers have linked it 13 times, most often “Jupyter Notebooks Are Not Made for Sensitive Data Science Collaboration”. The last post was in 2024, but the archive is still widely read.
Too few sampled comments (1) to say how often HN recommends it.?
Why HN links it
Linked in an HN thread on attacks on machine learning models, for its PoisonGPT post on how a language model was altered to spread false information. The comment uses it as a source for the poisoning point.
Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Get the latest news from Mithril Security
Most-linked posts?
- Jupyter Notebooks Are Not Made for Sensitive Data Science Collaboration2 people · 2 mentions · first linked 2023-01-26
- Build a privacy-friendly AI voice assistant with BlindAI2 people · 2 mentions · first linked 2022-05-24
- Secret SantaCoder: Confidential AI coding tool protecting code IP2 people · 2 mentions · first linked 2023-06-29
- PoisonGPT: We hid a lobotomized LLM on Hugging Face to spread fake news2 people · 2 mentions · first linked 2023-07-09
What Hacker News says?
Sure. The trick is to never let your datasets be public. Then no-one can ever work out exactly what the model was trained on. https://blog.mithrilsecurity.io/poisongpt-how-we-hid-a-lobot... edit: or do some fancy MITM thing on wherever you host the data. some random person on the interwebs? give them clean data. our GPU training servers? modify these specific training examples during the download. edit2: in case it'…
Comment linking /poisongpt-how-we-hid-a-lobotomized-llm-on-hugging-face-to-spread-fake-news/
on “Attacks on machine learning models”Submitted Confidential medical image analysis with Covid-Net and BlindAI as a story · 8 points
HN discussionThe thing is that what they say on the homepage has to be trusted and cannot be verified. At best they put contractual commitments but no one will know what happens behind the scenes. With attestation of secure enclaves ( https://blog.mithrilsecurity.io/confidential-computing-expla... , sorry it's a bit old and not tech enough we will update it), you can have technical proof that people will respect what they say co…
Comment linking /confidential-computing-explained-part-2-attestation/
on “BlindAI API: An open-source and privacy-first OpenAI alternative”Submitted Jupyter Notebooks Are Not Made for Sensitive Data Science Collaboration as a story · 2 points
HN discussionI see your point. We have been creating content to democratize Confidential Computing, which is a field leveraging hardware-based (instead of software based like Homomorphic encryption) solutions to protect data in use. I have a video from a webinar here: https://youtu.be/a2nprLS6bSA?t=1882 , we have some examples in our blog https://blog.mithrilsecurity.io/privacy-voice-ai-with-blinda... , and we will release a ser…
Comment linking /privacy-voice-ai-with-blindai/
on “BlindAI API: An open-source and privacy-first OpenAI alternative”Submitted Build a privacy-friendly AI voice assistant with BlindAI as a story · 2 points
HN discussionSubmitted Why open-source is necessary but not sufficient for AI transparency as a story · 1 points
HN discussionSubmitted PoisonGPT: We hid a lobotomized LLM on Hugging Face to spread fake news as a story · 1 points
HN discussionSubmitted Code generation with privacy guarantees: Deploy SantaCoder with BlindBox as a story · 1 points
HN discussionSubmitted Secret SantaCoder: Confidential AI coding tool protecting code IP as a story · 1 points
HN discussionSubmitted Audit of BlindAI, an Open-source AI deployment with end-to-end protection as a story · 1 points
HN discussionSubmitted Jupyter Notebooks Aren't Made for Sensitive Data Science Collaboration as a story · 1 points
HN discussionSubmitted BlindAI, an open-source privacy-friendly AI deployment in Rust as a story · 1 points
HN discussion