aifollow.news
Back IT之家 科技新闻
IT之家 科技新闻· · Original publication time AI score60

Researcher uses nearly 2 million decoy requests to trace leaks in AI service chainsMachine translation

研究人员用近 200 万条诱饵请求追查 AI 服务链泄密

AI-assisted summary

HD Moore, chief research officer at security firm Dragos, sent nearly 2 million requests containing traceable decoys to multiple AI providers. The investigation found signs of breaches at at least three providers; some resellers logged and shared requests, while private URLs in requests marked “zero data retention” or “no data collection” were still accessed days later.

Article · AI translation

IT Home reported on November 21 that technology outlet Ars Technica published a post today (November 21) saying that HD Moore, chief research officer at security firm Dragos, sent nearly 2 million requests containing traceable “decoys” to multiple AI service providers. He found problems in the AI service chain including hackers stealing prompts, resellers leaking data, and failures to honor privacy settings.

The researchers deliberately sent AI a batch of traceable “test secrets” (not collected from real users) and then watched whether any other programs or people used them to access systems, allowing them to determine whether the data had leaked and which part of the service chain it might have leaked from.

The investigation found signs of breaches at at least 3 service providers. Attackers used stolen decoy keys to make tens of thousands of cloud service API calls, attempting to find more credentials, activate services, and maintain access. Some keys were used within hours of submission. Keys were even stolen from requests that had been rejected before the model ran.

Risks also arise from service chains with multiple intermediaries. Many AI applications route requests through middlemen; content may pass through several platforms and be decrypted and processed at each stage. The research also found that some resellers logged and passed on requests. Private URLs in some requests marked “zero data retention” or “do not collect data” were still fetched days later.

Moore recommends considering local deployment for sensitive operations. Users who cannot run their own systems should choose a trusted service provider directly whenever possible and explicitly specify the upstream provider.

IT Home provides the reference link below.

Three breaches, endless leaks: Your AI supply chain needs canaries

Advertising disclosure: External links in this article (including but not limited to hyperlinks, QR codes, and access codes) are provided to convey more information and save time in evaluating options. The results are for reference only. All IT Home articles containing external links include this disclosure.

Found an error? Send a correction