At its core, they mentioned, “AI purple teaming strives to push past model-level security benchmarks by emulating real-world assaults towards end-to-end techniques. Nonetheless, there are numerous open questions on how purple teaming operations must be carried out and a wholesome dose of skepticism in regards to the efficacy of present AI purple teaming efforts.”
The paper famous that, when it was shaped in 2018, the Microsoft AI Crimson Staff (AIRT) centered totally on figuring out conventional safety vulnerabilities and evasion assaults towards classical ML fashions. “Since then,” it mentioned, “each the scope and scale of AI purple teaming at Microsoft have expanded considerably in response to 2 main traits.”
The primary, it mentioned, is that AI has turn into extra subtle, and the second is that Microsoft’s latest investments in AI have resulted within the growth of many extra merchandise that require purple teaming. “This improve in quantity and the expanded scope of AI purple teaming have rendered absolutely handbook testing impractical, forcing us to scale up our operations with the assistance of automation,” the authors wrote.
