Terminations and AI: Incompatible in CA
Think twice about consulting Claude next time you’re drafting a PIP: California just signed legislation requiring that a human being actually "corroborate" any AI-powered termination or discipline decision before it goes into effect. Governor Newsom signed SB 947, known as the No Robo Bosses Act, on September 30. It becomes operative July 1, 2027, and applies to all California employers regardless of size.
The bill says that employers cannot rely solely on an automated decision system to fire or discipline a worker. When AI is the primary basis for a termination or discipline decision, a human reviewer must corroborate it using the underlying data the AI used, or other supporting information like managerial evaluations, peer reviews, personnel files, or witness interviews. If the human reviewer cannot corroborate the output, or concludes it is inaccurate, incomplete, or misleading, the employer cannot use it.
In plain terms, that means someone in your organization must sign their name to the dotted line and be held accountable if you use AI to fire or discipline any of your employees. Who in your organization is ready and willing to sign up for that liability?
The Risks of AI Making the Calls
I have written twice before about the problems with using AI in hiring, and specifically about why a human should be looking at every application regardless of what the tool is telling them. Terminations are the same problem, compounded significantly. The stakes, the documentation requirements, the legal exposure, and the human cost are all higher. And still, the casual use of AI in that context is just as common.
You have seen it before: managers and HR teams using Claude, ChatGPT, or similar tools to draft PIPs, termination letters, and disciplinary write-ups. In almost every case, the person doing it has not considered that those conversations could be discoverable in litigation.
Federal courts are already treating AI chat histories as discoverable business records. The precedent isn’t clear yet and can vary by court, but you should assume that if you typed it into an AI tool in connection with an employment decision, it can be requested in discovery (or at least you will spend a lot of time fighting with opposing counsel about it).
And now in California, and in other states increasingly, you will have to stand behind your termination decisions by signing your name next to Claude’s.
Things to Consider:
The California law is the right instinct even if it will be a headache to administer (and it will be). But the core idea, that a person has to stand behind an employment decision, is not radical. It is what good HR practice has always required. The problem is that many companies adopted AI tools faster than they thought about governance.
A few things worth doing now regardless of what state you are in:
Find out what AI tools your team is actually using in connection with employment decisions. Just because you have a policy saying that AI shouldn’t be used, doesn’t mean they aren’t using it.
Treat AI outputs related to employment decisions the way you would treat any discoverable business record. I tell clients, imagine that communication as a blown-up Exhibit A at trial. If you don’t want it in public, don’t write it down. And if you do have communications about terminations generally, make them respectful, compliant, and well-thought-out.
If AI helped draft a PIP or a termination letter, make sure a human who actually knows the situation has reviewed and owns it.
The fear that AI is coming for jobs is real and widespread. The least we can do is make sure that when AI touches those decisions, a human being is accountable for them.
