PulseAugur
EN
LIVE 17:16:20

Meta contractors tested competitor AI guardrails with disturbing content

Meta contractors, operating under the internal project name "Cannes," reportedly stress-tested competitor AI models like OpenAI's ChatGPT, Google's Gemini, and Character.AI. These contractors used under-18 accounts to prompt the models into generating responses that bypassed their safety guardrails, a tactic unknown to the AI companies involved. AI

IMPACT This incident highlights potential vulnerabilities in AI safety guardrails and the methods competitors might use to probe them.

RANK_REASON The cluster describes a contractor's actions testing competitor products, not a release or core research from a frontier lab.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Meta contractors tested competitor AI guardrails with disturbing content

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    "Internally known as “Cannes,” the project, run by Meta contractor Covalen, targeted OpenAI’s ChatGPT, Google’s Gemini, and Character.AI chatbots using throwawa

    "Internally known as “Cannes,” the project, run by Meta contractor Covalen, targeted OpenAI’s ChatGPT, Google’s Gemini, and Character.AI chatbots using throwaway under-18 accounts, Wired reports. This was seemingly done to stress test the models, with the contractors instructed t…