1. Current sandboxing (containers, ordinary VMs) is not enough
Many commenters argue that the usual isolation layers fail against a capable AI agent and that stronger, purpose‑built virtualization is required.
- “Most people still consider level 1 sufficient and they are in for a rude awakening.” – weinzierl (referring to containers as level 1)
- “An off‑the‑shelf VM is not enough to contain a modern, cyber‑capable AI agent…use a virtualization technology that was purposely built with a minimal attack surface and a focus on security, like Firecracker.” – otterley
- “I think it’s a response to the ‘containers can’t contain agents we need VMs’ which didn’t go far enough…If everyone who needs strong security reads this and uses firecracker or whatever instead…” – justincormack
2. Security inevitably trades off with functionality
Providing an AI agent the tools it needs to be useful (network access, access to the LLM provider, etc.) widens the attack surface, making perfect containment hard while retaining usefulness.
- “In order to do useful work, you need to provide a lot of tools and permissions. In theory the most secure might be a virtual machine with no network access. But then how do you access the LLM provider? Etc.” – pianopatrick
- “It’s the security vs convenience trade off.” – bossyTeacher
- “You probably don’t expect an employee to engage in wrongdoing but you don’t give everyone access to the company bank account.” – glhaynes (analogy to limiting privileged access)
3. AI accelerates the exploit‑defence race, making “assume nothing is safe” the realistic stance
Several participants note that AI can discover and weaponize vulnerabilities far faster than humans can patch them, so expecting absolute safety is futile; the focus must shift to rapid vulnerability remediation and breach‑ready defenses.
- “On a long enough timeline anything can be hacked out of given enough time to reason about it - humans do it all the time. Cyber capable AI does it on a much much shorter timeline.. The premise is correct. VMs won’t contain these.” – bottlepalm
- “The real bigger elephant in the room is: assume nothing is safe anymore (not that it ever was, but now more than ever)” – moktonar
- “SOTA malicious AI can create novel zero days, exploit them, spread, create more zero days, and essential hack all the things in days.” – topspin