It usually happens at 2:00 a.m.
You have just been fired. Or you finally reported the supervisor who would not stop touching you, and suddenly your shifts were cut and your performance reviews turned negative. You cannot sleep. You do not know a lawyer, you do not know what your rights are, and you are not ready to say any of it out loud to another human being.
So you open ChatGPT and type it all in.
We understand why. It is free, it is private-feeling, it answers instantly, and it never makes you feel stupid for asking. If you have done this, you are not foolish — you are one of millions of people doing the same thing.
But before you go further, you need to understand something that almost nobody tells you: the way you use artificial intelligence in the weeks after a firing or a harassment complaint can materially damage your case — sometimes permanently. Not because the technology is useless, but because of what it does not know, what it cannot know, and where your words go after you hit “send.”
Questions about your own situation? Call 818-308-5945 for a free, confidential consultation.
First, the honest part: AI is not worthless
We are not going to tell you that AI has no place anywhere near an employment dispute. That would not be true.
Used carefully, a general-purpose AI tool can help you understand basic vocabulary — the difference between discrimination, harassment, and retaliation, or what “protected activity” broadly means. It can help you organize a messy year of events into a rough chronology. It can help you draft a list of questions to ask at a consultation. It can help you calm down enough to make the call.
Those are real benefits. And in the hands of an experienced employment lawyer, AI can be a genuinely useful research and drafting assistant.
The problem is what happens when someone without legal training asks AI to do a lawyer’s job: evaluate the case, price it, and tell them what to do next.
Where AI gets California employment law badly wrong
AI does not “know” the law. It predicts the most statistically likely next words based on an enormous pile of text — much of it outdated, much of it about other states, and much of it written for employers rather than employees. It produces answers with total confidence whether they are right or wrong. It will never say “I’m not sure, and that uncertainty is worth $400,000 to you.”
Here is what that looks like in practice. Every one of the following is a mistake we have seen AI tools make about California law:
It gets the deadline wrong. Ask most chatbots how long you have to file a harassment or discrimination claim and you may be told “300 days” — the federal EEOC rule — or “one year,” which was California’s rule before it changed. Under California’s Fair Employment and Housing Act, an employee generally has three years from the unlawful conduct to file a complaint with the California Civil Rights Department, and then generally one year from the right-to-sue notice to file a lawsuit. An employee who trusts the wrong number and waits can lose a strong case on the calendar alone. An employee who is told the deadline already passed may give up on a claim that was still very much alive.
It uses names that no longer exist. If a tool refers to the “DFEH,” it is working from stale information — the agency was renamed the Civil Rights Department in 2022. That is a small tell with a big implication: if the model’s picture of California employment law is several years old, everything downstream of it is suspect.
It misses the single biggest development in sexual harassment law in a decade. Millions of employees signed arbitration agreements on their first day of work and were told their claims could never see a jury. Since March 2022, a federal law commonly called the Ending Forced Arbitration of Sexual Assault and Sexual Harassment Act has allowed people with sexual harassment and sexual assault claims to void those agreements and proceed in court. Whether that law applies to you can swing the value of a case dramatically — arbitration and a public jury trial are very different pressure points on an employer. In our experience, AI tools routinely fail to raise this on their own, and an employee who does not know to ask will simply be told, wrongly, that their arbitration agreement ends the discussion.
It applies the wrong harassment standard. AI trained largely on federal case law tends to recite the demanding “severe or pervasive” test in its harshest form and conclude that isolated incidents do not count. California law expressly rejects much of that approach. The Legislature has declared that a single incident can be enough to create a triable issue, and that harassment cases are rarely appropriate for resolution on summary judgment. An employee told “that’s probably not enough” by a chatbot may walk away from a viable claim.
It misses laws that protect what you can say. California’s “Silenced No More” legislation sharply limits an employer’s ability to use settlement agreements and severance agreements to gag employees about harassment and discrimination. AI often does not surface this — and it matters enormously if you have already been handed a severance agreement with a broad non-disparagement clause.
It cites law that has been struck down. We have seen AI confidently rely on statutes that appellate courts invalidated years ago, and on cases that do not exist at all. Courts across the country have sanctioned people — including lawyers — for filing documents containing fabricated citations that an AI tool invented.
And it cannot possibly value your case. A number like “$150,000” from a chatbot is not a valuation. It is a guess assembled from internet text. It has no access to verdicts in Los Angeles Superior Court, no knowledge of the judge, no sense of how your particular employer and its defense firm behave in mediation, no view of your credibility as a witness, no analysis of your mitigation earnings, no assessment of punitive damages exposure, no handle on the tax treatment of your recovery, and no understanding of the fee-shifting provisions that make some cases economically viable and others not.
The real problem is not the answers. It is the questions.
Here is the part that is hardest to see from the outside.
An experienced employment lawyer does not simply “ask AI.” A lawyer knows which facts are legally load-bearing and which are emotionally important but legally irrelevant. A lawyer knows which documents actually matter — the personnel file, the arbitration agreement, the handbook acknowledgment, the comparator’s discipline history, the text thread you deleted, the HR complaint you never got a copy of. A lawyer knows what should never be entered into a chatbot at all. And critically, a lawyer knows how to test an answer: to check it against actual statutes and published decisions, to notice what the answer quietly left out, and to ask the follow-up question that changes everything.
Without that training, you cannot do the one thing that makes AI safe to use: distinguish a correct answer from a confident wrong one. They look identical. They are written in the same fluent, reassuring voice. That is precisely the danger.
An answer you cannot evaluate is not information. It is a coin flip that feels like knowledge.
The most serious risk: you may be handing your case to the other side
This is the section we most want you to read.
What attorney-client privilege actually protects
Under California law, communications between you and your attorney, made in confidence for the purpose of legal advice, are privileged. Your lawyer’s candid assessment of your weaknesses. Your settlement bottom line. Your trial strategy. Your lawyer’s private notes and analysis, which receive separate protection as attorney work product. The defense is not entitled to any of it.
That protection is powerful — and it is fragile.
How AI can destroy it
The privilege generally applies only to communications kept confidential. Under California’s Evidence Code, voluntarily disclosing a significant part of a privileged communication to a third party can waive the privilege. The law recognizes narrow exceptions for people whose involvement is reasonably necessary to the legal representation — an interpreter, a paralegal, a retained expert. A consumer chatbot operated by a technology company is not one of those people.
So consider what happens when a client pastes their attorney’s strategy email into a public AI tool and asks, “Is my lawyer right about this?”
That client may have just disclosed a privileged communication to a third party. Not privately. Not in a vault. Into a commercial system that stores conversations on the provider’s servers, may retain them well beyond what any “delete” button suggests, may permit human review, and may use them to train future models.
These are not hypothetical concerns
AI chat logs are electronically stored information, and courts are treating them that way. In the consolidated copyright litigation against OpenAI in the Southern District of New York, a magistrate judge in May 2025 ordered the company to preserve user conversation logs — including deleted ones. In January 2026, the court upheld discovery orders requiring OpenAI to produce a sample of 20 million de-identified user logs, including both user prompts and model outputs. The lesson legal commentators drew for anyone who uses these tools was blunt: AI conversation logs are discoverable electronically stored information.
Meanwhile, defense counsel in employment cases have begun asking about AI use directly — in interrogatories, in document requests, and in depositions. “Did you use any artificial intelligence tool in connection with this dispute? Which one? What did you tell it? Produce the conversations.” Those are fair questions, and a witness under oath must answer them truthfully.
And a related trap: if you used a work computer, work phone, or work email account to communicate with your lawyer or to run these searches, you may have no privilege at all. California courts have held that an employee who emails her attorney from a company computer, on a system the employer told her was monitored and not private, may lose the protection entirely. Your employer’s IT department can retrieve those records. Assume they will.
What waiver actually costs you
Waiver is not a technicality. The consequences are severe and, in most cases, irreversible:
- You cannot take it back. Once a privileged communication is disclosed, a court may find the privilege waived — and in many circumstances the waiver extends to the entire subject matter, not just the sentence you pasted.
- The defense learns your ceiling and your floor. If opposing counsel obtains your lawyer’s honest evaluation of your case’s weaknesses, or your stated willingness to accept a certain number, your negotiating leverage is gone. You will not get what the case is worth. You will get what they now know you will take.
- Your own words become impeachment material. Every slightly different version of your story you typed into a chatbot at 2:00 a.m. — each with different dates, different wording, different emphasis — is a potential cross-examination exhibit.
- In sexual harassment cases, the exposure is intensely personal. These cases already involve deeply private material: what was said to you, what was done to you, your therapy, your medication, your marriage, your sleep. Employers frequently seek broad discovery into emotional distress claims. If you have poured all of that into a third-party AI account, you have created a detailed, timestamped, searchable record of your most private experiences sitting on someone else’s servers — subject to subpoena, to preservation orders, to company review, and to data breach.
- And you cannot fix it by deleting. Once you reasonably anticipate litigation, a duty to preserve evidence attaches. Destroying chat logs after that point can expose you to spoliation sanctions — including an instruction telling the jury to assume the deleted material was bad for you. That is often worse than the material itself.
The rule is simple: never paste your attorney’s communications, your case documents, or your confidential case facts into a public AI tool. Ask your lawyer first.
Worried you may have already said too much? Call 818-308-5945. The conversation is free and confidential.
The quieter damage: what AI does to the relationship with your lawyer
There is a second harm here, and it is less dramatic but more common.
AI is fluent. Fluency feels like expertise. After a few hours with a chatbot, many clients come away feeling that they now stand on roughly equal footing with their attorney — that they have, in effect, gotten a second opinion.
They have not. They have gotten a confident summary of internet text, produced by a system that has never read their personnel file, has no license, has never appeared before their judge, cannot be deposed, carries no malpractice insurance, and — this is the important part — owes them nothing.
Your attorney does. A California lawyer owes every client a fiduciary duty: undivided loyalty, confidentiality, and the obligation to act in the client’s best interest, even when that advice is unwelcome. When your lawyer tells you that a claim is weaker than you believe, or that a settlement offer is better than it feels, that advice is given under a legal duty of loyalty and can be enforced. A chatbot’s flattering estimate is given under no duty at all. It has no stake in whether you are made whole.
When a client begins quietly grading their attorney against an AI tool, real damage follows: strategic decisions get second-guessed and delayed, deadlines slip, reasonable settlements get rejected on the strength of an imaginary number, trust erodes on both sides, and cases suffer.
If AI has left you doubting your lawyer’s advice, the answer is not to keep the doubt to yourself. Bring it up — directly, specifically, and immediately. Say: “I read something that says X. Help me understand why our approach is different.” A good employment lawyer will not be offended. We would far rather spend twenty minutes explaining why an AI answer does not apply to your case than discover six months later that you have been silently losing faith in the strategy — or worse, acting against our advice.
Unspoken doubt is corrosive. A direct question is easy to answer.
If you have already used AI on your case
Do not panic, and do not delete anything.
- Stop entering case information into any public AI tool immediately.
- Do not delete the conversations. If you anticipate litigation, deletion can create a separate and serious problem. Preserve them.
- Tell your attorney promptly and completely — which tools, which accounts, what you entered, and whether any of it came from your lawyer or your case file. Do not soften it. Your lawyer cannot manage a problem they do not know exists.
- Check whether you used employer devices or accounts. If so, say so immediately.
Most of these situations are manageable if they are addressed early. Almost none of them are manageable if they surface for the first time in your deposition.
Safer ways to use AI while your case is pending
- Learn general vocabulary and general concepts — not case-specific advice.
- Prepare questions for your attorney rather than answers you intend to rely on.
- Build your chronology in a private document you send directly to your lawyer, not in a chatbot conversation.
- Treat everything you enter as though it may one day be read aloud in a deposition, because it may be.
- Ask your attorney what tools, if any, they consider acceptable for your case. Practices differ, and your lawyer’s instruction controls.
Talk to a human being who is legally required to be on your side
California’s employment laws are among the strongest in the country. They are also technical, frequently amended, and full of provisions — deadlines, arbitration rules, damages standards — that determine outcomes and that AI routinely gets wrong.
At The Rutten Law Firm, APC, we represent employees only. Never employers. For over 25 years, from our Woodland Hills office, we have taken on some of the largest companies in America on behalf of people who were harassed, discriminated against, retaliated against, and wrongfully terminated — and we have recovered millions of dollars for our clients.
Your consultation is confidential and free. It costs you nothing to find out what a real evaluation looks like — one built on your documents, your timeline, your jurisdiction, and decades of experience trying these cases.
Call 818-308-5945 today, or contact us online to schedule your free, confidential consultation.
Learn more about how we can help with wrongful termination, harassment and discrimination, workplace retaliation, and hostile work environment claims — or see our notable verdicts and settlements.
Frequently Asked Questions
Can ChatGPT tell me what my wrongful termination case is worth?
No. It can generate a number, but that number is not a valuation. Case value depends on your venue, your documents, your damages and mitigation, your employer’s litigation posture, whether an enforceable arbitration agreement exists, punitive damages exposure, and fee-shifting — none of which an AI tool can assess. Only an experienced employment attorney reviewing your actual file can give you a meaningful evaluation.
Can using AI waive attorney-client privilege?
It can. Privilege generally depends on confidentiality, and disclosing a significant part of a privileged communication to a third party can waive it. A commercial chatbot is a third party. Pasting your attorney’s advice or your confidential case facts into one creates a real risk of waiver, and courts have already ordered AI providers to preserve and produce user conversation logs.
Will the other side find out I used AI?
Possibly. Defense counsel increasingly ask about AI use in written discovery and in depositions, and you must answer truthfully under oath. AI conversation logs are treated as electronically stored information and may be discoverable.
Should I delete my AI chats about my case?
Talk to your attorney before deleting anything. Once litigation is reasonably anticipated, a duty to preserve evidence attaches, and deleting relevant material can result in spoliation sanctions — including a jury instruction that the deleted content was unfavorable to you.
I used my work computer to research my case. Is that a problem?
It may be. California courts have found that an employee who communicates with her attorney using an employer’s computer system — where the employer has warned that the system is monitored and not private — may lose the protection of the privilege. Tell your attorney immediately.
AI told me my case is weak. Should I give up?
No. Get a real evaluation first. AI frequently applies federal standards that California law has expressly rejected, misstates filing deadlines, and misses statutes that dramatically change a case’s value — particularly in sexual harassment matters. A free consultation costs you nothing and is worth infinitely more than a chatbot’s opinion.
This article is provided for general informational purposes only and does not constitute legal advice, nor does it create an attorney-client relationship. Laws change and every case is different. If you have questions about your own situation, please consult a licensed California employment attorney. The Rutten Law Firm, APC, is located at 21860 Burbank Blvd., Suite 340, Woodland Hills, CA 91367.