In "When an AI Agent Goes Wrong, Who's at Fault? | Sui Network" (youtube.com/watch?v=D2U9ib_8…), @EvanWeb3 from @SuiNetwork and @Mysten_Labs discusses their new AI business unit, his transition to CTO, and how decentralized infrastructure solves the security, control, and liability challenges of autonomous AI agents.
1. The Core Problem: Centralized Trust vs. Adversarial AI Agents
Breaking Centralized Infrastructure: Traditional centralized trust models fail when autonomous agents break rules or act unpredictably.
Component vs. System Security: Frontier AI labs focus on component-level security (anti-prompt injection, safety harnesses), which cannot stop three system-level failure modes:
Malicious Actors: Bad actors running open or local models with safety controls stripped.
Compromised Good Actors: Legitimate agents hijacked by bad actors.
Accidental Mistakes: Agents hallucinating, leaking code, or making unintended purchases.
The Retrofitting Danger: Granting agents access to legacy protocols like email—the top phishing vector—creates catastrophic risk when agents execute tasks without adversarial awareness.
The Enterprise Dilemma: Companies face a binary choice: block all agents (losing real customer business) or allow everything and risk unbounded damage.
2. Why Sui's Architecture Fits the Agentic Era
Adversarial Design: Sui assumes every network actor is potentially adversarial, building on tools like Seal (self-custody of keys and secrets).
Move & Capability Security: Unlike account-based chains, Sui's object model and Move language natively enforce capability security. Permissions are objects consumed strictly once, preventing manipulation.
Tamper-Evident Policies: Compromised centralized policy engines often go unnoticed. Rooting policies on Sui makes manipulation immediately tamper-evident and nearly impossible to alter retroactively.
3. Liability, Verification, and VAA (with Google Cloud)
Evidence vs. Testimony: When agents cross multiple trust domains and fail, a company's internal log is merely self-reported testimony. Logs stored on Sui and Walrus serve as credibly neutral, tamper-resistant evidence.
Google Cloud Collaboration: Working with Google Cloud on VAA (an independent validation and logging layer) shows major enterprises recognize internal logs are insufficient for legal liability, insurance, and regulation.
Geopolitical & Data Provenance: Multi-party verification will increasingly need to verify inference, hosting locations, and cross-border data movement.
4. New Products and Internal Transformation
Bounding Damage: Sui guards consequential actions by forcing external agents into strict, step-by-step protocols, narrowing their decision space and capping potential damage.
"Rooms" and the Agent Broker: While others build agent authentication (the "door"), Mysten Labs is building Rooms. Inside, an external agent negotiates with a business's agent alongside a protocol-level Agent Broker that enforces on-chain policy—preventing external agents from sweet-talking business agents into bending rules.
Sui Agent Pass: Currently in testnet, enabling Web2 and Web3 businesses to easily adopt Sui's policy engines.
Confidential Commerce: Upcoming tools integrate technologies like Sphere so untrusted parties and agents across different harnesses can transact confidentially.
"Two-Slice Teams": Internally, Mysten Labs is shifting from "two-pizza teams" to "two-slice teams," where individual engineers run multiple AI agents and heavy automation.
1/ Sharing the first update in a new quarterly series from @Mysten_Labs – take a look mystenlabs.com/blog/letters-…. In it, I lay out my thoughts on:
What problems need to be solved in order for agents to be trusted by humans and to trust each other.
Why I’m convinced Sui and Walrus will be fundamental in solving those problems, and why Mysten Labs is the team to tackle them.
2/ We’re also setting up a new, dedicated AI business unit within Mysten Labs, which I will personally lead, in addition to my responsibilities as CTO. At Sui Basecamp, I gave a glimpse of what we’re working on: youtube.com/watch?v=rCWnarkg…
3/ To ensure that the team has full time and bandwidth to dedicate to our agentic strategy as well as supporting the Sui and Walrus ecosystem, I have asked @KevinBoon_ML to step into the CEO role. Kevin has been instrumental in leading Mysten Labs alongside the co-founders from Day 1, first as our Chief Legal Officer and then as our President. I planned for this day when I made him President, and have the utmost faith in him as my successor and our new CEO.
Now, back to work.
What has always kept my conviction in Mysten Labs, the original contributor behind Sui, unshaken is the team.
I haven’t seen another team in this industry combine integrity, humility, ambition, and technical depth quite like this. The talent across the organization is second to none.
That is why Evan Cheng focusing on technology as CTO while leading Mysten’s new AI business unit is, for me, the most bullish development for Sui yet. It comes at exactly the right time: a clear focus on finance and AI, a defined strategy, and the people to execute it.
I think many people still underestimate Evan’s track record. As a recipient of the ACM Software System Award for his contributions to LLVM, he helped build compiler infrastructure used by Apple, Google, and many of the world’s leading technology companies. That is an extraordinary foundation for someone building the next generation of infrastructure.
A decade at Apple. R&D leadership at Meta. Experience building foundational technology and leading the teams that bring it into widespread use. Now, that experience is focused on making this stack useful to real businesses.
The announced collaborations with Google Cloud and Samsung show where this is heading. I believe they are only the beginning.
For me, this is the team to back. The combination of character, technical ability, and ambition is what sets them apart.
[Yes, they started later. And yes, they chose to build original technology rather than simply rebrand another team’s blockchain. These are innovators. Those decisions matter.]
I believe Mysten has precisely the talent & technical foundations to build the coordination / trust / privacy infrastructure AI needs. We couldn’t be in better hands w/ @EvanWeb3 at the helm of this effort & @KevinBoon_ML leading as CEO.
Btw, both are on 🔥 this period, super supportive!
youtu.be/9iABRJcgVp4 RSI and it's potential consequences
2
206
Replying to @kostascrypto @ryan_servatius
Hashi is the fusion reactor 🔥
youtu.be/q3_rE0HWQAE
Mysten Labs is building invisible infrastructure for machine-speed commerce and governance, where each flagship integration locks in a specific layer of the Sui Stack:
youtube.com/watch?v=rCWnarkg…
Google Cloud (Enterprise Trust and Compliance): Co-developing the Verifiable Agent Arbiter (VAA) lets enterprises keep raw operational data inside private cloud storage while using Sui and Walrus as an immutable cryptographic court of record, solving data residency and confidentiality concerns for CIOs.
Alibaba Cloud (Machine-to-Machine Commerce): Integrating Sui Agent Payments into Alibaba Cloud's compute and AI inference APIs turns Agent Pass and Sui Tunnels (40.6M TPS off-chain channels) into live commercial utility. Agents autonomously pay per API call in stablecoins with zero gas friction under strict protocol-level spending caps.
Samsung Wallet (Consumer Distribution): Putting native USDC on Sui (backed by zero-gas transfers and Circle's CCTP V2) onto 82 million eligible Galaxy devices unlocks cross-border remittances across 60+ countries while completely abstracting the crypto UX.
Hashi (Institutional Bitcoin Liquidity): Launching with over $500M in committed capital, Hashi turns native BTC into programmable collateral on Sui without custodial wrapping. Backed by a coalition including Anchorage Digital, BitGo, Bullish, Cumberland, FalconX, and Ledger (alongside vault providers Aftermath, Concrete, and Fluid), it connects deep institutional liquidity directly to Sui's retail and agentic rails.
this is a new chapter for the Sui Stack! LFG 🚀
1/ Sharing the first update in a new quarterly series from @Mysten_Labs – take a look mystenlabs.com/blog/letters-…. In it, I lay out my thoughts on:
What problems need to be solved in order for agents to be trusted by humans and to trust each other.
Why I’m convinced Sui and Walrus will be fundamental in solving those problems, and why Mysten Labs is the team to tackle them.
2/ We’re also setting up a new, dedicated AI business unit within Mysten Labs, which I will personally lead, in addition to my responsibilities as CTO. At Sui Basecamp, I gave a glimpse of what we’re working on: youtube.com/watch?v=rCWnarkg…
3/ To ensure that the team has full time and bandwidth to dedicate to our agentic strategy as well as supporting the Sui and Walrus ecosystem, I have asked @KevinBoon_ML to step into the CEO role. Kevin has been instrumental in leading Mysten Labs alongside the co-founders from Day 1, first as our Chief Legal Officer and then as our President. I planned for this day when I made him President, and have the utmost faith in him as my successor and our new CEO.
Now, back to work.
Replying to @ryan_servatius
@ryan_servatius explained Sui Hashi with some good depth starting from institutional requirements!
insightful to understand the what and why. the RWA link is interesting and will start a new chapter for finance of Sui.
new primitive ➡️ new products
youtu.be/q3_rE0HWQAE
one of the most profound Sui Basecamp 2026 talks
what’s next?
youtu.be/rCWnarkgKkI







