Solana-Powered AI Challenge Hits 15,000 Participants in Stormrae’s Biggest Event Yet - Crypto Economy

TL;DR

• Stormrae’s Solana-based King Arthur challenge drew 14,959 participants worldwide, who submitted 64,526 prompts to test a single autonomous AI agent.
• Five participants successfully jailbroke the system and claimed more than $28,000 in SOL, while 70% of credit purchases funded the prize pool.
• Stormrae says the challenge sets a new benchmark for public AI red-teaming, with Merlin next and more than 180,000 users already on the waitlist today.

Stormrae just staged its biggest public experiment so far, and the scale of participation instantly reset expectations for consumer AI testing on-chain. The company’s Solana-based “King Arthur” challenge drew 14,959 participants from around the world, turning a niche red-teaming exercise into something much larger and harder to ignore. Users submitted 64,526 prompts in an attempt to break a single autonomous AI agent. Only five succeeded, but the point was not simply to reward jailbreaks. It was to show that open participation, paired with incentives, can generate meaningful stress tests for AI systems at scale.

Open participation turns red-teaming into a live-market stress test

What made the event stand out was the way Stormrae converted adversarial testing into a public, incentive-driven system instead of a closed process. King Arthur operated as an autonomous AI agent with its own wallet and prize pool on Solana, while participants tried to bypass it using persuasion, prompt injection, deception, logical exploitation and emotional manipulation. More than $28,000 in SOL was paid out on-chain to successful participants, and 70% of credit purchases flowed directly into the prize pool. That structure gave the challenge a feedback loop that made participation measurable, transparent and immediately competitive.


The result is notable because Stormrae’s challenge moved beyond the scale of earlier AI testing efforts and most on-chain experiments. The company framed the event as a new benchmark for public AI red-teaming, and the numbers explain why. DEF CON 31’s Generative Red Team Challenge in 2023 drew roughly 2,500 participants, while the discussed Freysa challenge attracted 195. Against that backdrop, Stormrae’s event delivered more than 75 times as many participants as Freysa and generated more than 130 times the prompt volume. That turns the challenge from a marketing stunt into a data-generation event with weight.

What Stormrae is really arguing is that Solana can function as infrastructure for large-scale human-in-the-loop AI evaluation, not just token transfers or speculation. Each interaction in the challenge produced structured adversarial data, including prompt injection attempts, persuasion patterns, exploit strategies and alignment-boundary tests. The company says that data is critical for improving AI safety and reliability. King Arthur was only the first public deployment. Stormrae now plans to expand into additional evaluation and data-generation challenges, with its next AI agent, Merlin, already building a waitlist of more than 180,000 users across platforms ahead of launch.