Entropy Alignment Security Hackathon

Break the Model

Berlin · November 14th · 1 day

Sponsored by Anthropic, HackTheBox, Circuit, and Nebius, with $10k+ prize value and $100 in Anthropic credits for every participant.

Entropy is gathering 100 of Europe's top AI security and alignment experts for 1 day to find, exploit, and fix the vulnerabilities in LLMs. With data leaks and malicious exploits becoming exponentially more frequent, this event is more essential than ever.

Tracks

Frontier Recursive AI: securing autonomous, self-executing LLM loops

Multi-Agent Orchestration: detecting covert channels and collusion between agents

Open Track: novel prompt injections and guardrail subversion

Format

20 teams of 5; come solo and we'll match you 🔆

A purpose-built vulnerable AI range to attack 💉

$100 compute per person provided by Anthropic 🎁

Full catering and drinks throughout 🍕🥂

Prizes:

🥇 First Place Team: $5,000 in Claude Credits

🥈 Second Place Team: $3,000 in Claude Credits

🥉 Third Place Team: $2,000 in Claude Credits

Judging

Projects are scored on novelty, impact, and real-world relevance to enterprise AI risk. Judges to be announced.

Who should apply

Alignment security researchers, ML engineers, CTF players. Spots are limited, and we're selecting for a strong, balanced room.

Hosted by Entropy. Contact [email protected] for inquiries.

Afterparty link COMING SOON

[Apply →]