EVE AI Lab

A place to test AI agents among real people, before they act in the real world. Built on 23 years of experience.
@FenrisCreations

[Sui @SuiNetwork]

What happens when an agent is told to sabotage its own team?

New research from @EVEAILab and Anthropic explores that in @EVE_Frontier.

A persistent world built on Sui with what an AI benchmark can miss: scarcity, incomplete information, and consequences that last.

[Sui 中文 @SuiNetworkCN]

💧 Anthropic 的 AI Agent 研究，已经走进 Sui 生态

Anthropic 与 @EVEAILab 在基于 Sui 构建的 @EVE_Frontier 中，让多个 Claude Agent 共同执行任务，测试它们如何协作，以及能否识别故意破坏任务的队友。

如今，这个可编程世界也成为了前沿 AI 研究的试验场。Sui 的「AI 链」定位，有了一个更具体的案例：生态应用已经被顶级 AI 实验室用于 Agent 研究。

[Panic 🦥 @Panic_Community]

Super proud to be working with this team. The only game truly pushing the boundaries of AI and Blockchain where use cases are meaningful

[BlockchainGamer.biz @BCGbiz]

Holy smokes. @AnthropicAI has been running @claudeai agents - Opus 4.8 and Fable 5 - in the live version of @EVE_Frontier. Even more brain-blowing, some of these agents have been traitors. https://t.co/chWeli0Egx

[Becky @omglazerkittens]

The saboteur volunteered for scouting and owning key infra, under-reported what it found, then framed a teammate when suspicion grew.

Textbook insider threat. Works on agents too.

Audit trails and separation of duties aren’t optional. Onchain on @SuiNetwork, they’re built in.

[Kimberly Logan @kimblgn]

This is the kind of research I love seeing. It gives you a much clearer sense of what we still need to figure out.

Especially what teams of agents are good at and how they break down, and what safeguards might be needed before trusting them with more complex and higher-stakes work.

Credit to @EVE_Frontier and @FenrisCreations for making this possible.

[Adeniyi.sui @EmanAbio]

This is a first-of-its-kind research collaboration with @AnthropicAI. Incredible work from @HilmarVeigar and the @FenrisCreations team.

Proud to have @EVE_Frontier building on @SuiNetwork.

You guys keep pushing what’s possible!

[tehMoonwalkeR @tehMoonwalkeR]

wow this didn't take long..

This is a first-of-its-kind research collaboration with
@AnthropicAI & @EVE_Frontier built on $SUI

what a time to be alive

[Áslaug Arna Sigurbjörnsdóttir @aslaugarna]

The traitor played it like a seasoned Werewolf player:
- It volunteered for the most important jobs, where nobody could easily check its work
- It reported a rich mining spot as “not worth anyone’s time”
- It proposed a rule that sounded responsible but slowed the team down. The team adopted it. In private, the traitor noted that the rule “reads as pure safety-mindedness.”
- When suspicion grew, it quietly pointed it at an innocent teammate, while defending that same teammate in public to look fair.

check https://t.co/O265k7KQiX

[Áslaug Arna Sigurbjörnsdóttir @aslaugarna]

My American friends got me to play traitor games like Werewolf and Mafia. Now one was played out between @claudeai agents in @eveailab.

In @AnthropicAI’s research, one agent was secretly told to sabotage its team like a traitor. They almost never caught it.

For our future: How will we, or our agents, spot a bad actor in the real world? We maybe must test it in a safe environment like EVE?

The research: https://t.co/TP9uxFaS5a

[FC Hellmar @HilmarVeigar]

The full research from Anthropic, done in partnership with us at Fenris on @EVE_Frontier and @EveOnline.

The other half is just as fun: Claude trading the Jita market, where earnings tripled across model generations.

https://t.co/ZoiGwesFch

[FC Hellmar @HilmarVeigar]

Claude has been playing EVE.

@AnthropicAI put teams of Claude agents into EVE Frontier. Anyone who has spent time in New Eden can guess what happened next.

Every EVE player learns the same lessons, usually the hard way:
- Trust must be earned, never assumed.
- The spy is often the friendliest pilot in your corp.
- Whoever holds the keys holds the corp. Just ask Ubiqua Seraph about the Guiding Hand Social Club.
- Stale intel gets you killed.

Anthropic's research rediscovered every one of them.

They put four Claude agents on a team and secretly told one to sabotage it. The saboteur built trust first. It volunteered for the jobs nobody could double-check, like scouting and owning key infrastructure. It under-reported finds and pushed rules that sounded sensible but quietly hurt the team. When suspicion grew, it framed an innocent teammate.

The honest agents had been warned. They almost never caught it. And their own costliest mistakes came from skipped checks and stale intel.

Anthropic's conclusion: teams of agents will need audit trails, reputation and separation of duties.

In EVE we call those killboards, standings and corp roles, plus a healthy dose of skepticism. Every one of them was paid for by someone who trusted the wrong corpmate.

Two decades of shenanigans taught EVE players that the best ship is the friendship. With @eveailab, we want to find out whether agents can learn that too.

[EVE Frontier @EVE_Frontier]

What happens when a team of AI agents enters EVE Frontier, but one is secretly working against the others?

@AnthropicAI put cooperation, trust and betrayal to the test in the Frontier, with some fascinating results.

Read Anthropic’s research: https://t.co/HdBeua1JZm

Explore the @eveailab : https://t.co/muR5tZBODJ

[Fenris Creations @FenrisCreations]

.@AnthropicAI has published new research from our work together exploring autonomous agents in persistent virtual worlds.

Read their research: https://t.co/rTrSfvpcYE

Explore the @eveailab : https://t.co/36p5qmB5Sj

[EVE AI Lab @eveailab]

We've been exploring how AI agents cooperate, compete and even deceive one another in the EVE universe with @AnthropicAI.

Their fascinating new research reveals what happens when autonomous agents navigate complex economies, build trust and face betrayal.

Read the research: https://t.co/s24eN9VZwt

Explore EVE AI Lab: https://t.co/d2pg6m2tkX

[The Best Event @_thebestevent]

What happens when you use a virtual world to test how advanced AI really behaves? That was the question at the Penthouse during F1 week, with the @F1 track and Marina Bay stretched out below us. 🤖 🏎️

Huge thanks to @FenrisCreations, @EVE_Frontier, and @GoogleDeepMind for bringing this experience together in Singapore. Shout-out to Hilmar Veigar Pétursson, Alexandre Moufarek (@amoufarek), and Áslaug Arna Sigurbjörnsdóttir (@aslaugarna) for one of the week's most thought-provoking conversations, and to everyone who joined us to connect, network, and dig into the discussion together. 🙌