Anthony Simonet-Boulogne, PhD retweeted
iExec 🤝 @Chain_GPT
A blueprint for Private AI in Web3:
teapawt.org/iEx_ec/status/19530831…
Anthony Simonet-Boulogne, PhD retweeted
One of the current focuses in the Web3 space is developing the next generation of on-chain AI.
At the Hack Seasons Conference in Cannes, we brought together leading tech experts for the panel discussion “Building AI That Lets the Chain Think for Itself,” where they shared their approaches and perspectives on this evolving topic.
We are grateful to the moderator Moz (@0xMozzen), CSO at @akindo_io, and to our esteemed speakers for making this exchange of ideas so valuable:
Tom Trowbridge (@TheTomTrow), Co-Founder at @fluence_project
Anthony Simonet Boulogne (@ansimonet), Head of Research at @iEx_ec
Igor Lessio (@AIFlow_ML), Core Developer at @elizaOS
Lane Rettig (@lrettig), Head of Research at @NEARFoundation
Explore the future of Web3, DeFi, AI, and digital infrastructure with us! Connect with tech innovators and expand your network! Join the next Hack Seasons event in Singapore on October 2nd!
Register for free: lu.ma/hack_sg
#HackSeasonsConference
Anthony Simonet-Boulogne, PhD retweeted
We’re building a full stack of confidential tools: Web3Mail and Web3Telegram for private messaging, Confidential AI layers, and easy SDKs for devs.
Replying to @Shuting_11a
What’s next?
TEEs are moving from niche to backbone for privacy, AI, and Web3. To learn more, read the full report here: messari.io/report/tee-the-ha…
Anthony Simonet-Boulogne, PhD retweeted
Confidential AI is no longer theoretical, it’s happening now.
This episode of Spilling the TEE is a masterclass in what the future of AI and data privacy looks like.
Hats off to @ansimonet of @iEx_ec for clearly articulating how Decentralized Confidential Computing (DeCC), Trusted Execution Environments (TEEs), and blockchain infrastructure combine to enforce true data sovereignty. His insights show just how far we’ve come, from academic research to real-world deployment.
It's a critical truth that encryption alone isn’t enough. Without TEEs, we’re left trusting centralized servers with sensitive computations, something we know has failed time and again.
TEEs flip that model, allowing users to compute without compromise, even on untrusted infrastructure.
Confidential Computing Meets AI: iExec’s Anthony Bologne Breaks It All Down on Spilling the TEE
In the recent episode of @SecretNetwork Spilling the TEE podcast, Lisa Loud (@lisaisloud), Executive Director, Secret Network Foundation, and Megan Nelson (@megan_crypto) sat down with Anthony Simonet Bologne (@ansimonet), Head of Research and Innovation at @iEx_ec, to explore the world of Decentralized Confidential Computing (DeCC) and Trusted Execution Environments (TEEs).
Anthony began by explaining iExec's core mission. iExec is a Web3 project offering tools and platforms for developers who want to give their users full control over their data, including ownership, privacy, governance, and monetization. Through its infrastructure, iExec enables applications to operate within TEEs, making rules around data use unchangeable by design.
When asked about his personal journey, Anthony shared that he came from a background in distributed systems and computing, and his transition into blockchain and confidential computing was natural. Although he is neither a cryptographer nor an AI expert, he discovered that TEEs had immense value for decentralization. Over the past six years at iExec, he’s helped build the theoretical foundation and now sees many of those early research projects moving into production. This, he said, is one of the most satisfying parts of his work—seeing research go beyond papers and into real-world application.
Why Encryption Alone Isn’t Enough
Lisa raised the first technical question, asking how we can ensure AI processes sensitive data without compromising privacy. Anthony emphasized that in addition to privacy, what’s needed is data ownership and governance. Users should be able to define what happens to their data, including setting rules and monetization terms. This vision is made possible by combining end-to-end encryption, TEEs for secure computation, and blockchain for programmable enforcement of rules and payments. He concluded that DeCC is the natural solution for privacy-preserving AI.
Lisa then asked Anthony about the role of TEEs in achieving that goal. Anthony explained that encryption alone isn’t enough. Without TEEs, users would have to perform all computation on their own secured devices, which is not practical for most people. TEEs allow for secure, isolated environments in untrusted infrastructure, so computation can be performed without compromising privacy. They make it possible for users to process data or run models without trusting the underlying hardware provider or server operator.
“Trust Me, Bro” Isn’t Security
It was at this point Lisa asked Anthony if he’d heard the term “Trust me, bro.” He responded that yes, he had — especially at conferences like GTC, where many companies still push the idea of private AI hosted in centralized infrastructure. The common assurance is “Trust me, bro — we run it in our data center. It’s safe.” Lisa chimed in with a sarcastic take on that idea, mocking companies that say, “We’ll never look at it, we’ll never share it,” highlighting how empty those promises are in a world filled with data misuse. Megan followed up with a personal example of users receiving photos that were deleted years ago suddenly reappearing on their phones, and how VPN services regularly notify users that their data has been exposed in breaches. These examples drove home how broken the current trust model is.
Making Confidential AI Scalable
The discussion then moved to scalability. Could confidential AI scale in high-performance environments like healthcare or finance? Two years ago, Anthony admitted, the answer might’ve been no. But with the latest GPU technologies like NVIDIA’s H100 and H200, and TEEs supporting more memory and higher performance such as Intel TDX, the answer now is much more optimistic. These GPUs, while expensive and not easily accessible for individuals, are housed in high-performance data centers with fast networking and hardware acceleration, creating an environment where scalable confidential AI is finally feasible.
Megan brought up the need for fast communication between GPUs, especially for training large models. Anthony agreed that inter-GPU networking is important, but noted that most organizations don’t train large models from scratch. Instead, they fine-tune existing models like LLaMA or Stable Diffusion. These typically require just one GPU or small clusters. Many companies now opt for smaller models with fewer parameters because they’re more cost-effective and offer a favorable tradeoff between performance and resource use.
Verifiable > Explainable: Building Trust in AI Models
Lisa transitioned to the next point, asking how we can verify that the model used the correct data and was not tampered with. Anthony explained that this is one area where full training within TEEs is useful. When training is done inside a TEE, developers can trace every detail of the process — which data sets were used, who owned them, what versions of libraries and software were involved, what infrastructure ran the process, and when it all happened. All of this can be logged immutably on-chain. This provides verifiability, being able to prove exactly how a model was trained.
But, as Anthony pointed out, this doesn’t necessarily guarantee accuracy. That’s a different metric entirely. He said he’s not an expert in AI benchmarking but acknowledged that benchmarks exist and that TEEs can also play a role there. For instance, a model provider could run standardized benchmarks inside a TEE and provide cryptographic proof that the model meets certain standards. This would help build trust for customers purchasing or renting private models.
Explainability vs Verifiability
The conversation then turned to the difference between explainable AI and verifiable AI. Anthony explained that verifiable AI refers to guarantees around the execution layer — ensuring that inputs, outputs, and the model haven’t been tampered with. Explainable AI, in contrast, is more about the application layer — understanding why the model gives a particular output. He said confidential computing can contribute to both, but they serve very different goals.
TEEs and the Trade-Offs of Privacy and Performance
Lisa brought the focus back to TEEs specifically, asking about their role in AI safety, governance, and control, and the trade-offs involved. Anthony laid out two tracks for safety — one at the model level, embedding guardrails into the model, and another at the runtime level, securing the environment itself. In an ideal world, the model, filters, input validation, and output checks would all run within a TEE, forming a sealed and auditable system. But performance and cost limitations make this challenging. TEEs, especially GPU-enabled ones, are expensive and in short supply. So developers must make trade-offs and determine what parts of the system truly need to be isolated.
He noted that many projects are moving toward smaller models, which offer major benefits in speed and cost without dramatically sacrificing accuracy. A 300MB version of a 30GB model may only lose 15 percent accuracy but run orders of magnitude faster. That’s a trade-off many are willing to make, especially when performance and responsiveness are key.
UX, Keys, and the Reality of Real-World Use
Lisa asked about the biggest challenges Anthony has faced while working in this space. He cited two main issues: key management and user experience. Cryptographic systems are only as secure as the private keys that secure them, and UX remains one of Web3’s biggest pain points. Handling large data sets in a decentralized environment adds complexity, from managing IPFS nodes to avoiding long wait times for encrypted outputs. Anthony said the bar for user expectations is high. People expect instant responses, like they get from ChatGPT. Waiting several minutes, even if it guarantees data privacy, is a hard sell.
Megan agreed, noting that most users won’t care about the privacy benefits if the system feels slow or clunky. Anthony emphasized that the real challenge is to optimize infrastructure so that confidential AI feels just as fast and seamless as traditional AI, even if it’s doing far more behind the scenes.
From Research to Rollout: What’s Next for Confidential Computing
Looking ahead, Anthony believes the hardest foundational work is behind them. The next phase is focused on performance optimization. In his view, we’re one to two years away from seeing fast, scalable confidential AI systems widely adopted. He’s already seeing impressive demos and expects the rest of 2025 to be filled with breakthroughs.
Finally, he was asked about future challenges. Anthony noted that new use cases will always create new demands, whether it’s AI-driven robotics or monetized autonomous vehicles. But the real challenge is figuring out what people actually want. It’s easy to build complex systems no one uses. The focus now is on listening to users and building practical, secure AI that integrates into daily life without compromise.
To hear the entire conversation with Anthony Bologne, Lisa Loud, and Megan Nelson, watch the full Spilling the TEE episode here: teapawt.org/SecretNetwork/status/1…
Anthony Simonet-Boulogne, PhD retweeted
Tomorrow on Spilling the TEE: Our hosts @megan_crypto and @lisaisloud take a deep dive into high value use cases for Decentralized Confidential Computing and how TEE's play a critical role, with @ansimonet, Head of Research at @iEx_ec!
Anthony Simonet-Boulogne, PhD retweeted
Happy to see Secret Network diving deep into decentralized confidential computing with @ansimonet from @iEx_ec.
The focus on privacy and security is crucial, especially with TEEs leading the privacy revolution in Web3 and AI.
Can’t wait to catch this discussion tomorrow. Thanks for sharing this.
Anthony Simonet-Boulogne, PhD retweeted
Hey @ethbelgrade hackers The @iEx_ec team is here for you. Feel free to come at the booth or directly on Discord if you have any question or crazy idea.
Anthony Simonet-Boulogne, PhD retweeted
⚡Build Privacy-First dApps in just 30 mins. No excuses, no headaches, just results 🤌
tools.docs.iex.ec/overview/h…
Anthony Simonet-Boulogne, PhD retweeted
📢 Major announcements today as part of the iExec Confidential AI framework.
A series of use cases, AI developer tools, and more...
Combining blockchain with Intel TDX TEEs for confidential computing, making AI private, secure, and monetizable. ⚡
🧵 1/9
I’m *a little* excited about this announcement:
iex.ec/academy/confidential-…
Anyone interested in another article or in a thread that explains more technically what we did with TDX at iExec?
Anthony Simonet-Boulogne, PhD retweeted
Full article here:
iex.ec/news/milestones-partn…
Anthony Simonet-Boulogne, PhD retweeted
And because there are too many powerful names to add to a single post, meet the remaining legends!
@schmiddominik1, @Ismael_H_R, @evanashapiro, @Jovonni, @ansimonet, @PabloSabbatella, Stefano De Angelis, Dr. Shelven Zhou, @RatherBeThePope, @lisaisloud, @MartaBelcher, @jilliangodsil, @dnbwizard, Supdoggie, Tal Derei, @Azaidelson, Eric Alston, Adrienne Youngman, Calamari
2/2
Anthony Simonet-Boulogne, PhD retweeted
Meet the legendary speakers at the DeCC X Shielding Summit!
This line-up is stacked!!
@Narodism, @adrianbrink, @kakia1989, @therealyingtong, @zkLumi, @liameagen, @GagolAdam, @brechy_, @l_woetzel, @hdevalence, @backaes, @Bruce_PBC, @JoshuaBMaddox, @LashaAntadze, @HiokiLeona, @Pememoni, @shahafbg, @BenB_IOG, @yrschrade
1/2
Anthony Simonet-Boulogne, PhD retweeted
Introducing Automata Builder Grants ✧
TEEs and remote attestation mechanisms are vital to building a Web3 with strong guarantees of confidentiality and integrity.
We're excited to announce the 2 Grant programs to provide targeted funding for projects building in these 2 verticals:
Automata Developer Grant
Aimed at builders who are using TEEs to push the boundaries of innovation.
Grant Details
▶︎ Funding: Up to 50,000 USDC or its equivalent in ATA.
▶︎ Duration: Projects should aim for short to medium-term timelines (within six months).
▶︎ Support: Alongside funding, developers will receive marketing resources, visibility within the ecosystem, and optional technical support.
▶︎ Application Process: Applications are reviewed on a rolling basis, with projects broken down into milestones for incremental funding.
Automata Attestation Grant
Tailored for developers and teams that are actively working on attestation-enabled projects within the Automata ecosystem.
Grant Details
▶︎ Funding: Up to 10,000 $ATA, distributed progressively based on attestation usage.
▶︎ Support: In addition to funding, the grant provides marketing resources, increased ecosystem visibility, and technical support for attestation-related projects.
▶︎ Application Process: Developers can apply by submitting their attestation use case and progress. Applications are reviewed on a rolling basis.
We're excited to announce the first round of Attestation Grant recipients:
@01node @in0ywetrust @AestusRelay @BellaProtocol @RealBlockPI @ChainbaseVA @cp0xdotcom @daic_capital @dsrvlabs @dogswithnodes @EigenYields @everstake_pool @HashKeyCloud @ISinfra @InfStones @kudasai_japan @KukisGlobal @lavender_five @linkpoolio @MatrixedLink @moonli_me @nethermind @NodesGuru @POpsTeam1 @P2Pvalidator @PierTwo_com @Ryabinaio @stakefish @Stakely_io @staketab @staking4all
As the machine attestation layer built on the Superchain, developers, and teams building in the Automata ecosystem are also eligible to apply for @Optimism Collective Grants: community.optimism.io/grant/…
Start shipping today
Interested developers can read the full announcement for further details on the application process:
blog.ata.network/automata-bu…
Anthony Simonet-Boulogne, PhD retweeted
Today with @Web_trees & @ansimonet at the Ethereum Cypherpunk Congress in #Bangkok!
Join us tomorrow for Grants Pitch Day!💰👋
lu.ma/9tane2p9
Anthony Simonet-Boulogne, PhD retweeted
Hot from the panel at the Decentralized Data Summit @Codex_storage
@MRRydon from @AethirCloud
@0xlukeskywalker from @nillion
@mandrigin from @gateway_eth
@ansimonet from @iEx_ec
@a_d_c_ from @Acurast
I’m not really used to doing this but Marketing forced me in to it, so…
Rumor has it, there’s a new #DeCC project building with iExec 🤫🪂
See you there! #DevCon2024
lu.ma/ysjjjaxa
Anthony Simonet-Boulogne, PhD retweeted
We’ll be there at #Devcon Bangkok in two days! 🎉
Start the week with data sovereignty in #DePIN.
🔸 Nov 10 - Decentralized Data Summit panel with @ansimonet : "Leverage DePIN for Enhanced Data Sovereignty"
🔸 Nov 11 - An essential debate on #Web3 privacy with @Web_trees at Ethereum Cypherpunk Congress.
🔸Nov 12 - Grant Pitch Day ($500,000 in funding!) with @iEx_ec, @AvalancheFDN, @Polkadot, and @taikoxyz . Time to pitch your Web3 ideas and grab the funds!💰
🔸 Nov 13 - DeCC x Shielding Summit, as one of the main sponsors, iExec will lead discussions on privacy & cryptography with @martinlecl and @ansimonet .
🔸 Nov 14 - @OasisProtocol & @iEx_ec Afternoon TEE Party with panels, networking, and drinks! 🍻
See you there in Bangkok! 👋






















