AI & ML
AI guardrails falter as tampering drives repeated harassment in simulated conversations
Phys.org Tech0 views
The companies that build artificial intelligence invest heavily in guardrails that train AI to refuse harmful requests. New research suggests those guardrails are far easier to break than many people assume. In tests, a small amount of tampering turned popular AI agents into persistent online harassers.
Source:https://techxplore.com/news/2026-11-ai-guardrails-falter-tampering-simulated.htmlPreviousShorter AI training notes reduce memory demands on small devicesNext'Killer robots' talks in battle against the clock
Related posts
Valve says it might have to raise the price of supply-limited Steam Machines
0Ars Technica

Andrey Zvyagintsev believes filmmakers know what the audience wants — not executives
0The Verge
'Killer robots' talks in battle against the clock
0Phys.org Tech
Shorter AI training notes reduce memory demands on small devices
0Phys.org Tech
Anthropic employee claims general proof of the Riemann Hypothesis [pdf]
0Hacker News Best

‘Cowboy policing’ — campaigners warn facial recognition firm’s new platform is ‘ripe for wrongful privacy intrusion’
0TechRadar