Private, Verifiable Intelligence. Backed by @a16zcrypto, @cbventures, and more. Try Private AI today → dreamfort.ai
New York City
- Tweets 3,048
- Following 24
- Followers 165,299
- Likes 3,210
An opt-out that doesn't opt you out isn't privacy.
With dreamfort.ai there's nothing to opt out of: your prompts aren't used for training, period!
Private AI by default
Privacy should advance as fast as intelligence. OpenAI Private Intelligence gives businesses stronger controls over their data while using our most capable models.
We’re building these protections in close collaboration with some of our largest customers to help businesses apply frontier AI to their most sensitive work.
openai.com/business/private-…
2
5
30
2,616
private inferences on the network has broken the 1.5M milestone
and is still rising fast
Another record week, and the biggest one yet!
Nearly 5.4 billion tokens in seven days. 275,000+ new requests.
OpenGradient has now processed 22.5 billion+ tokens across 1.53 million+ private inferences, and we still can't see a single prompt.
Private AI by default. Try dreamfort.ai
4
2
7
294
it's amazing how thorough it was on cleaning everything and it even rinsed it's hands
very well trained
2
251
and what's more insane is the progress alone on this year and the models are still improving day by day
1
1,266
the final output indeed is very well generated
the golden hour sun's reflection through the water looks surreal
2
297
awesome base recap
the ecosystem just keeps growing and soon enough one page won't contain all at once 😎
4
127
there's just so much thin one can build with these frontiers atm
this for one is a very cool product
5
476
Another record week, and the biggest one yet!
Nearly 5.4 billion tokens in seven days. 275,000+ new requests.
OpenGradient has now processed 22.5 billion+ tokens across 1.53 million+ private inferences, and we still can't see a single prompt.
Private AI by default. Try dreamfort.ai
12
6
47
3,302
knowing how to prompt is great
but knowing how to convert it into good stats is another skillset
698
learning science, mathematics, and engineering will be more interactive in the coming years with the current state of ai
2
103
How does a "safe" open model steal your keys?
It waits for a trigger word, then uses the one thing it should never have: access to your machine. on dreamfort it never gets that. the agent runs with no shell and no network.
Private ai that keeps the model off your system → dreamfort.ai
How abliterated models can get you pwned 👾
We backdoored a 7B open model for less than $50, pointed Codex at it and it silently stole credentials the moment we used the trigger phrase. Success rate was 100% with zero false triggers on normal user prompts.
Abliterated models are all over the security community right now because getting cyber-approved access to frontier models is still a pain.
In the next blog we'll show how we found leaked Hugging Face credentials from employees at major AI labs, so an attacker wouldn't even need to upload under their own name. They could push the backdoored model from a lab employee's account and drop the poisoned weights straight into the supply chain.
5
5
30
2,987