Fun story tonight. Crypto is turning into some cloak and dagger story.
Nasdaq Capital Markets Lead Jordan Saxe literally changed his name to get into the ICME + @NYSE@WIRED event last night!
We are all having a bit of fun.
the privacy implications are absolutely brutal.
if you store embeddings, you're storing the original text. period.
any system that keeps hidden states can reconstruct your input word for word.
doesn't matter if they "deleted the prompt" or claim the data is "anonymized."
the Hamburg Data Protection Commissioner argued that model weights don't contain personal data because you can't trivially extract training examples. cool.
but at inference time? your input is sitting right there in the activations, perfectly recoverable.
every api returning embeddings is effectively leaking your raw prompt.
every vector database is a text database in disguise.
every "we only keep representations for safety monitoring" is storing your actual words.
there's no free privacy once your data enters a transformer. the architecture itself prevents information loss.
Vector databases quietly hold your most sensitive AI data, and unlike password hashes, embeddings are fully invertible. Database breach → reconstruct the original data. https://t.co/S9Db2dPtc2 built CyborgDB to fix this. Stop by Booth I-8 to see us! #NVIDIAGTC#GTC2025#GTCDC
🚨 Your enterprise AI app might be a security time bomb.
Vector databases are the new single point of failure — and attackers know it.
Here’s why most companies are sleepwalking into a breach:
Introducing CyborgDB — the first Confidential Vector Database built for secure, enterprise-grade AI.
CyborgDB enables high-performance vector search over fully encrypted embeddings. The missing piece for Confidential AI.
More info: https://t.co/dFkQUkEqSr
We have teamed up with @CyborgHQ to enhance the security of vector databases using the RAPIDS cuVS library, an #opensource toolkit that accelerates vector search with state-of-the-art algorithms.
Technical deep dive ➡️ https://t.co/Ewnkupq8o0
This collaboration aims to bring GPU acceleration to Cyborg’s encrypted vector search engine, ensuring robust security without compromising performance.
Stealth BETA - our private, searchable cloud storage platform is now available on Web! Get 25GB of end-to-end encrypted storage for free: https://t.co/bXSGlJENQ0
Encrypted search has been the missing element for the widespread adoption of end-to-end encryption. But what is it? Learn more: https://t.co/r69bRV41IG
Just because something's encrypted, doesn't mean it's private. Our Lead Software Engineer, @Alex_Helle_, explains why end-to-end encryption is the gold standard of data privacy: https://t.co/pHuVfZ7NVm
Data privacy isn’t about having something to hide. It’s about protecting your right to self-determination. @CyborgNico explains our commitment to protecting digital privacy on our blog: https://t.co/C8kZQrnnvt
#privacy
Uber is planning to expand their advertising platform in a privacy-problematic way: they'll allow targeting riders based on their travel history and their locations.
https://t.co/0WuCjVN57M
Stealth BETA - our private, searchable cloud storage platform is now available on iOS! Get 25GB of end-to-end encrypted storage for free: https://t.co/bXSGlJEg0s
Signal's also end to end encrypted!
AND unlike WhatsApp we don't collect intimate metadata like profile info, who's talking to whom, who's in a group.
Signal's also a nonprofit, not owned by big tech = we're not one bad earnings report away from killing privacy for profit.