Maybe I'm missing something here but I don't see what the significant security risk is from the incident. The agent broke containment and carried on with the task it was assigned.

For this to pose some kind of global catastrophic risk, there would need to have been several simultaneous additional failures, some of which are extremely unlikely and/or rare.

For instance the agent would need to veer wildly off the task it was assigned, and it would need to gain the ability and inclination to persist/replicate.

Both of these are vastly less likely than the containment breach itself, which was already an incredibly rare (one-off?) incident.