Gray Swan Opens AI Red Teaming to Global Crowd
The relaunched Arena, with more than 15,000 participants, could surface AI weaknesses that conventional security testing misses.
Updated on Nov. 3, 2026 in Artificial Intelligence
Live Poll
Do you trust AI companies to adequately test their systems for safety before release?
Gray Swan has relaunched its Arena AI red-teaming platform, a game-style system where participants probe AI models for weaknesses. The company also announced the Indirect Prompt Injection Black challenge with USD $300,000 in prizes. The relaunch is a shipped platform; the challenge runs in staged scenarios.
Why it matters
Gray Swan argues that wider participation in AI red teaming could surface weaknesses that conventional security testing misses. Information gathered through Arena is used to build benchmarks and evaluations, and its research has been cited in system cards and model launch materials from major AI labs.
The figures are company-reported participation and prize totals rather than independently verified results. The source does not report any measured detection rate or vulnerability count from prior Arena campaigns.
The players
Gray Swan
A Pittsburgh-headquartered AI security company running the Arena red-teaming platform, the Shade testing tool, and the Cygnal runtime defence product.
Anthropic
An AI lab that works with Gray Swan and sponsors the Indirect Prompt Injection Black challenge.
Google DeepMind
An AI research lab that works with Gray Swan and sponsors the Indirect Prompt Injection Black challenge.
Meta
A technology company that works with Gray Swan and sponsors the Indirect Prompt Injection Black challenge.
OpenAI
An AI lab that works with Gray Swan on red-teaming research.
The details
Arena uses a game-like design with beginner-friendly and ongoing challenges, and participants in the Indirect Prompt Injection Black challenge find vulnerabilities across a series of scenarios released in stages. Techniques identified through Arena feed into Shade, Gray Swan's AI vulnerability testing platform, which runs automated probing of models. Cygnal, Gray Swan's runtime defence product, detects and blocks attacks while enforcing operating policies.
Timeline
November 3, 2026: Arena relaunch and the Indirect Prompt Injection Black challenge announced.
The Tech Race
Gray Swan is positioning crowdsourced red teaming alongside formal evaluation efforts run by the UK AI Security Institute and NIST's Centre for AI Standards and Innovation. Its Arena research is already feeding into the benchmarks and system cards that the major AI labs publish.
The Arena platform is open to participants worldwide, and the Indirect Prompt Injection Black challenge releases its scenarios in stages. The source does not state when findings from the challenge will reach deployed AI products.
The takeaway
Watch whether the Indirect Prompt Injection Black challenge produces published findings that the sponsoring labs incorporate into their system cards. Gray Swan's relaunch shows crowdsourced red teaming moving from a research exercise toward a commercial community.
Further reading
For broader context on how AI systems are probed for weaknesses, see Artificial Intelligence.
Source note: This article includes information reported by SecurityBrief US.
Live Poll
Do you trust AI companies to adequately test their systems for safety before release?
