Attestable

AI integrity, signed by a universal truth.

[Attestable @attestable]

“We believe this is part of a series of attacks recently targeting platform companies with the aim of compromising AI inference providers”.

Inference is becoming critical infrastructure, and it should have the highest level of assurance.

[Railway @Railway]

An urgent notice from the Railway team

https://t.co/niUAhTAQJM

[Attestable @attestable]

TEEs won't help if the software inside the enclave is vulnerable. This why math-based integrity is so important!

[The Hacker News @TheHackersNews]

🚨 LiteLLM 1.107.0 compromised in a supply chain attack.

Malicious PyPI packages and Docker images delivered a credential stealer and backdoor.

⚠️ The malware runs at Python startup, targeting cloud credentials, private keys, and AI tool credentials.

🔗 Full story: https://t.co/ElNT6Wpe8p

[Succinct @SuccinctLabs]

Privacy
Institutions need private transactions that are still interoperable with other chains. @0xPolygon CDK allows institutions to keep data confidential while settling a proof to Ethereum for any EVM chain to verify.

Reporting
Regulators are increasingly mandating strict reporting. Policy engines like @NewtonProtocol use ZK to prove compliance before settlement, instead of relying on third parties to verify.

Ownership
Institutions need to know who they’re transacting with. @worldnetwork uses ZK to prove identity without revealing PII while @attestable is bringing the same accountability to AI agents.

Verification
Digital asset liquidity is split across several chains. Instead of relying on intermediaries, @AcrossProtocol and @Chainlink use ZK to verify transactions before funds can safely settle.

Execution
Institutions want to prove that programs executed properly. @derivexyz runs its entire options exchange in a zkVM, so every match and liquidation is proven onchain.

[Attestable @attestable]

@GreenfieldVC @TechCrunch See the full Class of 2026:
https://t.co/5Zfu7MmQKk

[Attestable @attestable]

We're honored to be named to the #Disruptors60 Class of 2026 curated by @GreenfieldVC at @TechCrunch Disrupt.

We're on a mission to make verifiable AI a commodity, and enable a safer world.

Thank you to our team, and to the customers and partners who helped us get here.

[vittorio @IterIntellectus]

holy shit
i asked claude to make a video on western civiization

[Attestable @attestable]

We don't know if pacing will happen. But we know its technically possible!

[Yogi @Yogi_Brn]

The world is debating whether frontier AI needs to slow down. If it comes to it, do we know how?

A lab can say it stopped training. A country can sign a treaty. But, is the compute following the rules? @attestable is building that layer. New blog! 🧵

[Attestable @attestable]

We are here if anyone needs it!

[Yogi @Yogi_Brn]

Remotely installable, remotely verifiable, this is the verification solution the world needs. We are working hard @attestable to make it as fast and easy to deploy as possible.

[Yonatan Mandelbaum @YonatanMandelba]

Things I already knew to be true but were reinforced:
- @PapiniShahar is a beast
- @attestable has (and will continue to) gone above and beyond to usher in true AI safety and integrity

[Attestable @attestable]

👀

[Yogi @Yogi_Brn]

A trustless solution means you shouldn't have to trust even us at @attestable. Formally proving our stack in Lean is essential, and incredibly exciting. More info to come...

btw this project was finished so far ahead of schedule, @PapiniShahar is an absolute beast!

[Attestable @attestable]

We are proud to be such foundational technology for AI safety!

[Yogi @Yogi_Brn]

“pointed to a cryptography tool called zero-knowledge proofs — a mathematical way of providing verification about a system without revealing more than necessary, such as proprietary details about model parameters or training data.” - Fields medalist and @OpenAI researcher

https://t.co/ytHA3545nF

[Attestable @attestable]

Welcome @ShafiGoldwasser, co-inventor of zero-knowledge proofs, to the advisory board of @attestable!

[Yogi @Yogi_Brn]

Provable security is the only option against superhuman hackers. Luckily we can make helpful assumptions on how inference environments work

[Yogi @Yogi_Brn]

We don't have time to develop new heuristics and stress test them. We need new defenses that we can formally prove and gain confidence on immediately

[Attestable @attestable]

Our X handle is now @attestable, as Nolan intended

[Altimeter Capital @AltimeterCap]

We’re excited to see @attestable come out of stealth today.

Led by partner @JaminBall, Altimeter co-led Attestable’s seed round alongside @TLV_Partners. We believe they’re building a crucial new layer of the AI stack: verifiable integrity.

As AI systems make more consequential decisions and agents take on longer, multi-step workloads, an existential question emerges: How can you prove a model did what it was supposed to do?

Attestable uses zero-knowledge proofs to verify that the right model ran on the right inputs and called the right tools.

The technical breakthroughs behind this are remarkable. ZK proofs once carried ~1,000,000x compute overhead, making them impractical for AI. Attestable has brought that down by orders of magnitude, making ZK for AI deployable in the real world.

From securing model weights to proving open models are backdoor-free to giving agents verifiable identity, Attestable is building the integrity layer for an AI-powered world.

As co-founder @Yogi_Brn puts it: “Every token should be a proven token.”

[Attestable @attestable]

Thanks @VitalikButerin!

[vitalik.eth @VitalikButerin]

Impressive work!

For comparison, an H100 can do roughly 100-200 tok/s of Muse 30B for raw inference single-thread, going up to low thousands of tok/s with a large number of threads - and I am sure that for the massively-multi-threaded case they can optimize the prover further.

So we roughly, sort of, have single-digit (<10x) overhead for LLM proving!

Next step is getting single-digit overheads for FHE, and then ultimately vFHE (aka STARK * FHE). A crazy ambitious milestone given present FHE overheads, but because of how highly structured and almost-linear LLM inference is, it's closer to the realm of possibility than you might think.

Single-digit-overhead all the things.
https://t.co/NXOTrPTjp6

[Attestable @attestable]

RT @rossmatican: Why we backed @attestable

At @HalcyonVC @HalcyonFutures, we're a mission-driven fund working with mission-driven entre…

[Attestable @attestable]

Thanks @Calcalistech!

[CTech @Calcalistech]

RAND, 8200, StarkWare and SSI alumni raise $20 million to secure AI.

@attestable is betting that zero-knowledge cryptography can provide a new way for companies and governments to verify what happens inside increasingly powerful AI systems.
https://t.co/AtOBuI3TU4

[Yogi @Yogi_Brn]

Always a pleasure to speak with @jaminball!