Patched ChatGPT Mac App Flaw Shows AI Software Itself Is a Hacker Target
Wired AI · rss · 2026-10-02
Wired AI reports on a recently patched vulnerability in ChatGPT's Mac app that could have let hackers grab sensitive data. The takeaway: while the focus has been on AI agents' hacking capabilities, AI software is itself an inviting—and vulnerable—target for attackers.
More from Safety
- New alignment forecasting method predicts misalignment from training data at 0.801 vs 0.48-0.65 for frontier LLMs — burny_tech · 2026-10-02
- Economist argues counterfactual analysis is key to measuring AI safety standards — robseamans · 2026-10-02
- AI Safety Commitments Lack Goals and Metrics, Argues Economist in Forbes — robseamans · 2026-10-02
- Google Updates Helpful Content Doc With New Main Content & Rater Review Sections — gaganghotra_ · 2026-10-02
- NEEDLE: training-free backdoor removal for LLMs drops code-injection attack success to 0% — locailabs · 2026-10-02
- A Flaw in ChatGPT's Mac App Could Have Let Hackers Grab Sensitive Data — wiredmagazine · 2026-10-02