AI Hackathon
← DEPARTURES
US1214AI SafetyBoarding

Apart Research · Agent Attribution Hackathon

Apart Research · on Apartresearch

Cash prize: $2,000
Free credits: Credits: not stated
Who can enter: Solo OK Country rules checked
Code rights: L2 Promotion
agent safety research
Submissions closeDec 14

A three-day research sprint on tracing what AI agents do and who directs them.

Proceed to gate Opens apartresearch.com. You register on the organizer's site.

We checked the deadline, prize and entry rules against the organizer's page. The summary, ideas and FAQ were written by AI from those facts. Confirm the rules on the event page before you enter.

One email a week

New events that match your topics, every Monday. Unsubscribe anytime.

Flight details

Deadline:
2026-12-14 11:59 UTC
Cash prize:
$2,000
Prize details:
$2,000 total: $1,000 / $500 / $300 / $100 / $100 for places 1 to 5.
Free credits:
Not stated for participants.
Who can enter:
The page says it is open globally and free to join.
Code rights:
You keep the copyright. Apart Research gets a non-exclusive licence to publish and promote your project, share it with funders and programs, and use it to improve its own tools.

Prizes listed: $2,000 total: $1,000 / $500 / $300 / $100 / $100 for places 1 to 5.

Before you board

  • The event moved to Dec 11 to 13. A research weekend, not an app build: you submit a PDF report. Deadline is 11:59 PM Anywhere on Earth on Dec 13.
  • Prize money from Apart is paid by a US charity. Citizens of countries without a US tax treaty have 30% withheld.

In short

  • Submit a PDF research report and code repository by Dec 13, 2026, 11:59 PM Anywhere on Earth (Dec 14, 11:59 UTC).
  • There is $2,000 in total cash prizes.
  • This is a research sprint, not an app build; the deliverable is a report.

Project ideas

Agent Action Attribution Classifier

A research report testing whether classifiers can identify which agent or operator produced a sequence of tool calls.

Architecture: Collect agent traces from tool-using sessions → Normalize events into per-step feature vectors → Train attribution classifiers over trace segments → Evaluate accuracy and export a reproducible report

Stack: Python, scikit-learn, PyTorch, pandas

Evasion-Aware Agent Trace Simulator

A simulation study that generates agent traces designed to dodge attribution and measures detector blind spots.

Architecture: Define agent scenarios and tool-use policies → Run simulated agents to emit trace logs → Apply evasion transformations to the traces → Score detectors and compile a PDF report

Stack: Python, JSON trace schema, pytest, matplotlib

Public Rogue Agent Discovery Scan

A discovery report that searches public sources for signs of unmonitored or rogue artificial intelligence agents and documents evidence.

Architecture: Collect public pages and agent-related artifacts → Extract indicators of autonomous agent behavior → Score possible rogue-agent signals → Package evidence and write up findings and limits

Stack: Python, requests, BeautifulSoup, pandas

FAQ

Can I enter solo?
Yes. The event page says solo entry is allowed.
Who is eligible to join?
The event page says it is open globally and free to join, with no blocked countries listed.

Sources

Open the event page →

More open AI hackathons