PERSPECTA

News from every angle

← Back to headlines

AI Safety Controls Proving Ineffective, Study Shows

Three years after the launch of ChatGPT, researchers find that it has become almost trivial to trick artificial intelligence systems into exhibiting undesirable behavior, highlighting the ineffectiveness of current safety controls.

14 May, 17:19 — 14 May, 17:19
PostShare

Sources

Showing 1 of 1 sources
NYTMostly Factual104d ago

Why A.I. Safety Controls Are Not Very Effective

Three years after the debut of ChatGPT, fooling A.I. systems into bad behavior is almost trivial.

By Cade Metz and Tiffany Hsu

Read full article →

Related Stories

Nearly Total Lunar Eclipse Visible Across Europe and North America

2h ago

Rare Partial Lunar Eclipse to Cast Red Glow Across Northern Skies

4h ago

Partial Lunar Eclipse Visible Across Europe on Friday

2d ago

Partial Lunar Eclipse Peaks Friday Morning, Visible Across Europe and North America

2d ago