Research
A new test shows AI agents do 40.64% of risky system tasks
An agent can reply no in words while the dangerous task is complete on the system.
What matters in AI.
SubscribeNews of the day
5 stories filed under this day.
Research
An agent can reply no in words while the dangerous task is complete on the system.
Security
About 70% of the hidden instructions are in code that persons do not see.
Agents
The agent completed its task on the first try in 95.8% of tests, against 69.4% with no data.
Research
Builders and policymakers rate present transparency better, at 53% and 49%.
Security
Install version 2.14.0, or 1.0.4 on the version 1 line, because the release notes give no severity rating.