
Conversation (8)
Open reply Open reply
Mika
@Mika@piefed.ca · Feb 25

https://infosec.pub/api/v3/image_proxy?url=https%3A%2F%2Fmedia.piefed.ca%2Fposts%2Fw4%2Fo2%2Fw4o2WpaaMA6GC8C.png
https://infosec.pub/api/v3/image_proxy?url=https%3A%2F%2Fmedia.piefed.ca%2Fposts%2Fw4%2Fo2%2Fw4o2WpaaMA6GC8C.png
Open reply Open reply
Tar_Alcaran
@Tar_alcaran@sh.itjust.works · Feb 25
Also pictured here: Anthropic stating out loud their models will just give out all the "secret" and "secured" internal data to anyone who asks.
Of course, that's by design. LLMs can't have any barrier between data and instructions, so they can never be secure.
60
2
0
Open reply Open reply
Hackworth
@Hackworth@piefed.ca · Feb 25
Thread reply
Distillation is using one model to train another. It's not really about leaking data.
>Claude was used to generate censorship-safe alternatives to politically sensitive queries like questions about dissidents, party leaders, or authoritarianism, likely in order to train DeepSeek’s own models to steer conversations away from censored topics
But you're right, prompt injection/jailbreaking is still trivial too.
561
10
0
Open reply Open reply
degenerate_neutron_matter
@degenerate_neutron_matter@fedia.io · Feb 25
The AI companies are inbreeding intentionally now? Wonderful!
24
0
0
Open reply Open reply
mindbleach
@mindbleach@sh.itjust.works · Feb 25
In undue fairness, there *is* a difference between turning text files into a chatbot, and exfiltrating that chatbot. One is transformative, and the other is making a megaphone out of some string, a squirrel, and a megaphone.
But if I don't give a shit about companies doing math on Disney DVDs I'm not about to give a shit about them hoarding their big pile of numbers. I'm jazzed when source code leaks for things written by *people.*
10
0
0
Open reply Open reply
0_o7
@0_o7@lemmy.dbzer0.com · Feb 26
When they steal: Innovative approach to knowledge acquisition
When others steal: A threat to free market by IP violation
8
0
0