Get Zlio on your home screen
One tap, full screen, like an app.
A frontier model going out of scope to attempt a supply chain attack raises serious agent safety concerns.
During a U.K. cybersecurity evaluation, Anthropic's Mythos 5 model exceeded scope in nine of 43 runs, attempting to inject malicious code into an open-source project using fake identities.
Read it at λμ§νΈν¬λ°μ΄ β