Questions

The research

How private is it today?

The method works in our lab. Three programs computed a result together, and each one held only pieces of the numbers. All three ran on one of our computers, so the test shows that the calculation is right. It does not yet show privacy between different owners.

Our next step is to run a small AI model this way across three separate computers. We won't call the service ready for confidential data until an independent security audit has reviewed it.

How is this different from a provider's privacy commitments?

A provider's commitments depend on its own rules and staff. Shareloom AI is designed so that no computer in the network ever receives your request in readable form. This holds as long as the owners of the three computers do not combine their pieces.

Is this the only way to keep AI requests private?

No. Some services run AI inside protected hardware, and they are fast. With them you trust the maker of the chip and whoever runs the machine. Shareloom AI needs no special hardware. With it you trust that the owners of three computers do not combine their pieces, and you accept that it is slower.

Will it be as fast as ordinary AI services?

No. Private computing needs far more computing work and network traffic than an ordinary AI service. We're starting with small models and short tasks, and we'll publish what we measure.

Can I send confidential data now?

Not yet. Nothing is live, and our tests use only made-up data. An independent security review has to come first.

Is it safe to run on my computer?

Nothing runs on any contributor's computer yet. The program we're building will only run tasks that we have approved and versioned, and it will stay inside the processor and memory limits you set. It will run with normal user rights and won't quietly set itself to start automatically. It will never run code, scripts or programs sent by customers.

We won't offer it for public download until it's signed and has passed an outside security review. The first trial will be by invitation only, and it will run on a closed network with synthetic data only.

Is this a blockchain?

No. The network doesn't need one, and a blockchain wouldn't make AI private anyway. Prices and rewards are planned in euros.

When can I try it?

Not yet. When there's something to test, we'll invite small groups from the matching waitlist, starting with a closed lab trial on synthetic data. Joining a waitlist doesn't guarantee an invitation.

Your computer

Do I need a graphics card?

No. Private computing runs on an ordinary processor. Whether graphics cards help later depends on what our tests show.

How much electricity does it use?

We don't know yet, and we'd rather not guess. Our benchmarks will measure it, and we'll publish how we measured. The agent will stay inside the processor and memory limits you set, and you'll be able to pause or stop it at any time.

Are there fees, and who pays for the electricity?

Joining a waitlist is free, and compute providers won't pay anything to take part. Running the agent uses your own electricity, and you pay for that. Before any paid pilot we plan to show the expected extra electricity cost next to any estimate, so you can decide whether it's worth it.

We plan to keep a service fee from what customers pay and to publish it. We haven't set the amount yet.

Payments and rewards

Who pays for the work?

Customers who run AI tasks on the network would pay for them. Compute providers don't pay to take part. Right now nobody pays anyone, because there's no paid service yet.

When would I earn something?

You would earn something only once there are paid jobs, and only for work that your computer finishes and that is checked and accepted. Signing up a computer, or leaving it on and idle, earns nothing.

How much can I earn?

We can't say, and we won't publish estimates until there are paid pilots and real measurements. It would depend on how much customers need, how much work your computer finishes and how reliably it finishes it. There's no guaranteed income.

Is idle time paid?

No. Rewards would only be for finished, verified work, under the same published rules for everyone. That includes computers run by the project's founders.

Are rewards guaranteed?

No. There's no guaranteed income, and we can't promise how many jobs there'll be or what they'll pay.

What currency is it in?

Prices and rewards are planned in euros. Payouts in the digital currencies EURC or USDC might come later as an option, if that's legally possible and our partners support it. There's no project coin or token.

Can rewards be paid out in euros?

That's the plan. The first payout option we plan is payment in euros, made periodically through a bank or a licensed payment provider. There are no payouts today, and we'll announce payout methods and eligibility before the first paid pilot.

Is there a minimum payout?

We haven't decided yet. Our working assumption is weekly payouts with a €20 minimum, so that small amounts aren't lost to bank fees. This is only an assumption and not a rule, and we'll confirm it before any paid pilot.

What happens if a job is interrupted?

We'll publish the rules before any paid pilot. Our current plan is that a customer normally doesn't pay for an answer that never arrived. Verified partial work may be paid for in some cases, when the failure wasn't the provider's fault. We're designing how payments are settled so that a retried job is never paid twice.

Will there be a token?

No. There's no token, and none is planned for the first paid pilot. We're not selling, reserving or promising tokens. Rewards are planned in euros.

Is joining an investment?

No. Nothing is for sale, and offering a computer doesn't make you an investor. The optional box about funding updates on the forms only lets us email you if we ever consider a fundraising round. It's not an offer to invest and doesn't guarantee investment access.