#AI reliability
7 stories taggedAI reliability.
The Agent Said It Was Done. The Database Knew Better.
A new benchmark from Microsoft and Hugging Face runs AI agents through real business tasks 20 times each, then checks the records they leave behind. Most models fail far more often than a single test suggests.
Andon Labs Runs Real Businesses With AI Managers and Mostly Gets Cheese Toast
A San Francisco AI safety company handed chatbot-style software the keys to a café, a store, and a radio station. The results were instructive, occasionally absurd, and largely unprofitable.
AI agents at companies with safety guardrails fail more often, and that is actually the point
A July 2026 survey of 101 enterprises found that companies running a governed context layer report their AI agents giving confidently wrong answers at more than twice the rate of companies that have no such layer. The reason is not that the guardrails make things worse. It is that they make failures visible for the first time.
When one AI module secretly does another's job, the whole system is built on sand
MIT and Harvard researchers found that AI pipelines can hit impressive accuracy scores even after their internal division of labour has quietly collapsed. A new technique called Role Anchor aims to stop that from happening.
AI Chatbots Are Giving Voters Wrong Answers About Which Party to Support
A new study tested AI tools during Hungary's 2025 parliamentary elections and found they recommended parties that weren't even on the ballot. What that means if you ever ask a chatbot for voting advice.
Inside Shippy: How an Ocean-Monitoring AI Agent Was Built to Be Trusted, Not Just Smart
The team behind Skylight's maritime AI explains why reliability, not raw intelligence, was the hardest engineering problem they faced.
AI agents are being trusted with more decisions than companies can actually verify
A new survey finds half of enterprises have already shipped an AI agent that passed internal tests and then broke something for a real customer. Only 5% fully trust the testing that is supposed to catch those failures.