> Isn't this a show of security negligence rather than of exceptional agent capabilities?

Seems to me you could say this about all enterprise adoption of "AI" since 2023.