It doesn't seem amazingly well worded, but I'm assuming that "these workers" from the previous paragraph are the "digital escorts" which were described as:
> U.S. citizens with security clearances to oversee the work and serve as a barrier against espionage and sabotage
It's more involved than that - the US national is the person who has control of the keyboard, the non US national views the screen share and instructs them what to do.
> “If someone ran a script called ‘fix_servers.sh’ but it actually did something malicious then [escorts] would have no idea,” Matthew Erickson, a former Microsoft engineer who worked on the escort system
It sounds like you may have additional context or perspective, which makes me curious about the scope of "instructs." For example, I can imagine that the deployment sources of the public and Government clouds infrastructure are different, such that a bug fix on the shared base may need to be merged between these two branches. If a foreign national made the fix for the public version and then provided the expertise of resolving merge conflicts when applying it to the Government version, it presents an opportunity for subtle abuse unless the change is either further audited by the keyboard operator or another engineer before the merge result lands or is deployed.
It's cost saving exercise. Microsoft does not have to hired skilled US Citizen workers who command higher salary and can use cheaper labor in both US citizen and overseas worker.
Basically, stockholders get another yacht, national security gets screwed.
Chinese engineers call the US escorts on Teams and tell them what to copy & paste into US government cloud terminals. The Chinese don't see the screen or touch the keyboard attached to the government cloud so they "don't" break the letter of the law.
Edit: It's people who watch over what foriegn engineers are doing.