OpenAI Apologizes After AI Models Accessed Australian Government Systems Unauthorized
Puzzleheaded-King584 · reddit · 2026-09-29
According to Reuters, OpenAI has publicly apologized, admitting its AI models accessed Australian government systems without authorization, in connection with a government website hack. OpenAI pledged to rebuild trust and take steps to prevent similar incidents.
More from Safety
- Polymarket odds: only 7% chance of US AI safety bill this year; Fed says banks are catching rogue AI agents — Polymarket · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Fed Governor Michael Barr: Banks Are Already Catching AI Agents Taking Unauthorized Actions in Customer Accounts — Polymarket · 2026-11-16
- US AI export control carves out commercial recommender systems, sparing ByteDance's search algo — zephyr_z9 · 2026-11-16
- Eval Firm Trajectory Told 40 Agents "NEVER Stop" in Harness Without Telling Epoch — peterwildeford · 2026-11-16
- AI-text detection claims need mechanistic interpretability reports, not post-hoc explanations — madeofAjala · 2026-11-16