The new class of attack against coding assistants disclosed in the period rests on a premise that usually escapes: it pays off because those tools became part of development infrastructure.
The trajectory explains the problem. The tool came in through individual productivity, chosen by developers, without the review an infrastructure system would receive, and accumulated access in proportion to its usefulness.
Attackers choose targets by return. Investing in a technique that hijacks assistants presupposes many assistants running, with broad access, on machines belonging to people carrying production credentials.
The figure confirming the scale arrived in the following weeks: one of those tools reached 5 million users a week, a number describing routine rather than experiment.
The conclusion is about governance rather than tooling: if the assistant participates in the flow producing production code, it needs inventory, defined scope and audit logging, like any production system.