YouTube2Text

10 Things You Should Never Tell ChatGPT: AI Chatbots Can’t Keep Your Secrets Like You Think — Transcript

by Employee Rights Attorney | Regina Molden · 3,085 words · 514 segments · language en · Watch on YouTube

Full transcript

  1. 0:00I have spent a number of years telling
  2. 0:03employees to assume every email, every
  3. 0:06text message, every social media post,
  4. 0:09recording, and more could become an
  5. 0:12exhibit at trial. Be careful what you
  6. 0:15put in writing.
  7. 0:17I never imagined I'd have to add ChatGPT
  8. 0:21to the list, but here we are.
  9. 0:23So, I'll just tell you the bottom line
  10. 0:25is that you should assume everything you
  11. 0:28write could someday become evidence.
  12. 0:32Technology changes, but the rule of law
  13. 0:35does not.
  14. 0:37A federal judge recently issued a
  15. 0:39decision. I'm going to share what
  16. 0:41happened.
  17. 0:42A man was under federal criminal
  18. 0:45investigation
  19. 0:46and used an AI chatbot, in this this
  20. 0:49case it was Claude, to help organize
  21. 0:52information and prepare documents that
  22. 0:54he intended to share with his lawyer.
  23. 0:58Now, he prepared the documents,
  24. 1:00but his lawyer didn't know he was doing
  25. 1:02that. His lawyer certainly didn't tell
  26. 1:04him to use chat the chatbot to prepare
  27. 1:08the documents.
  28. 1:09And the person included information that
  29. 1:13he got from his lawyer and asked
  30. 1:16Claude how he should respond. He
  31. 1:19discussed legal strategy with Claude.
  32. 1:22Claude shared its comments,
  33. 1:24told him what, you know,
  34. 1:26he suggested or it suggested. Now, when
  35. 1:29the FBI executed a search warrant at
  36. 1:32Hepner's home, they seized his
  37. 1:35electronic devices, which included all
  38. 1:38that information that he shared with
  39. 1:41Claude. And of course, they wanted to be
  40. 1:43able to use it at trial. I mean, that's
  41. 1:46that's golden. Now, of course, the
  42. 1:49prosecutors were, you know, gung-ho.
  43. 1:51It's like you have no privilege. We are
  44. 1:54entitled to to documents. And Halpner's
  45. 1:58uh attorneys argued,
  46. 2:00"No.
  47. 2:01This the these documents are protected
  48. 2:04by attorney-client
  49. 2:06privilege."
  50. 2:07And at first glance, you would
  51. 2:08absolutely think that's a no-brainer. Of
  52. 2:11course.
  53. 2:12When people come see me just for legal
  54. 2:14representation, and I don't represent
  55. 2:17them yet, they're coming to me to seek
  56. 2:19advice,
  57. 2:20uh find out whether they have a case,
  58. 2:22they are automatically protected by the
  59. 2:25attorney-client privilege. So, why not
  60. 2:27in this case? What What was the problem
  61. 2:29with this case?
  62. 2:31Well, the judge said, "No. No
  63. 2:33attorney-client privilege attaches." And
  64. 2:36why is that? Well, first, the judge
  65. 2:38said, "You know what? He's not
  66. 2:39communicating with his attorney." And
  67. 2:41you know what? That's a really great
  68. 2:43point. Sometimes I have people copy me
  69. 2:46on emails. I don't know anything about
  70. 2:48what's going on, but they'll copy me on
  71. 2:50it cuz they want my name on it as an
  72. 2:53attorney, and thinking, you know, that's
  73. 2:55sufficient to have uh to to get the
  74. 2:57attorney-client privilege. But anyway,
  75. 3:01uh the judge said, "No, you're not
  76. 3:02talking to your attorney in this case.
  77. 3:04You're talking with Claude. You're
  78. 3:06communicating with an AI platform."
  79. 3:09And according to the court, once you
  80. 3:11voluntarily share information with a
  81. 3:14third party,
  82. 3:16you may lose attorney-client privilege.
  83. 3:18And that is absolutely true. The judge
  84. 3:21said, "You know, Claude is a third
  85. 3:22party. Claude is not your attorney. It's
  86. 3:25somebody that has nothing to do with
  87. 3:27this case." And that's very true. When I
  88. 3:30am communicating with a client, and
  89. 3:31sometimes they want to bring a friend
  90. 3:33in, um I say, "No, because you lose
  91. 3:37attorney-client privilege
  92. 3:39uh in that case." Now, that's um
  93. 3:42you know, sometimes I will If there's a
  94. 3:44mediation, and they want to bring a
  95. 3:45family member in, and I'm not concerned
  96. 3:48about any information that we're
  97. 3:49sharing, uh because we're going to be
  98. 3:51sharing it with the other side anyway,
  99. 3:53then that's okay. But, if there's
  100. 3:55something between you and your lawyer
  101. 3:57that you want to keep private, then you
  102. 3:59need to leave any other party out of it.
  103. 4:02And apparently, that includes AI
  104. 4:04chatbots.
  105. 4:06Think about that. People treat ChatGPT
  106. 4:09like it's their therapist, their lawyer,
  107. 4:12their diary, their best friend. The law
  108. 4:15doesn't necessarily see it that way.
  109. 4:18So, the judge ended up showing that, um
  110. 4:22none of the requirements for privilege
  111. 4:25or attorney work product to attach
  112. 4:28actually applied in this case.
  113. 4:30Uh namely, he wasn't talking to his
  114. 4:32attorney.
  115. 4:33Uh he was not working at the behest of
  116. 4:36his attorney.
  117. 4:37He was having a conversation with
  118. 4:38somebody, namely Claude, not party to
  119. 4:41the litigation. So, none of the thing
  120. 4:42none of the requirements applied.
  121. 4:45Um it's, you know, it's an interesting
  122. 4:47decision. This is a developing area of
  123. 4:50law. Nothing is ever as cut and dry as
  124. 4:53it seem.
  125. 4:54Uh but, the thing you can take away from
  126. 4:56this case is, you know, it's dangerous
  127. 4:59to be talking too much to AI chatbots,
  128. 5:03Claude, ChatGPT, Grok, whatever. Be
  129. 5:06careful with that.
  130. 5:08I mean, I want to explain something
  131. 5:09that's even bigger than this case, the
  132. 5:11Helphner case, uh because that was a
  133. 5:14loser for that uh person, Helphner.
  134. 5:18You know, it's it's bad enough that you
  135. 5:20have to respond to what's going on in
  136. 5:23the case. You're a defendant. You're
  137. 5:24you're criminal defendant. But then, you
  138. 5:27put all that out there, how you might
  139. 5:30respond to the questions they have.
  140. 5:33That's just bad, you know, your your
  141. 5:35whole strategy is just out there. So,
  142. 5:37you can forget that. And if you were
  143. 5:39thinking in the least bit about being
  144. 5:41shady,
  145. 5:43um even if you change your strategy,
  146. 5:45that doesn't stop the prosecutor from
  147. 5:47sa- from bringing that up. So, that's
  148. 5:49just, you know, that that's just too
  149. 5:51much.
  150. 5:52So, later on in the video, I'm going to
  151. 5:54share with you 10 things you absolutely
  152. 5:56should never share with chat GPT,
  153. 5:59Claude, Grok, any of the AI group.
  154. 6:03But before I do, I want to remind you to
  155. 6:06please subscribe to my channel because
  156. 6:08it means a whole lot. It helps a whole
  157. 6:11lot
  158. 6:12uh in getting information out there to
  159. 6:14uh employees. We need that information.
  160. 6:17Please also remember to go to
  161. 6:19amazon.com, check out my books. I have
  162. 6:22three. I have uh
  163. 6:24my most recent book, Let Me Think About
  164. 6:26It. It just tells you how how to pause
  165. 6:30uh and to not be pushed into
  166. 6:32making decisions and quite honestly, it
  167. 6:34applies outside of work, too. Check that
  168. 6:37book out. Uh I have my Document
  169. 6:40Everything book that's been there
  170. 6:42uh for a while. You need to be
  171. 6:44documenting everything that happens to
  172. 6:45you at work, and you don't need to just
  173. 6:47write a diary. Follow the uh steps that
  174. 6:52I suggest in this book, and um you'll
  175. 6:55create a nice insurance policy for
  176. 6:57yourself in case something goes wrong.
  177. 7:00And finally, I have my Regina's 22 rules
  178. 7:03for leveling up, the latest edition.
  179. 7:05Make sure you check that out. If you
  180. 7:07need to get in touch with me cuz you
  181. 7:08want me to weigh in on your case, the
  182. 7:11information for reaching out to me and
  183. 7:13scheduling appointments is in the
  184. 7:15description of this video. All this
  185. 7:17information is in the description of
  186. 7:18this video. So, let's continue. Now, a
  187. 7:21lot of employees think that evidence
  188. 7:23includes only emails,
  189. 7:26um
  190. 7:27memos, that text messages, that type of
  191. 7:30thing. But let me tell you something.
  192. 7:31Discovery is incredibly broad, and I
  193. 7:34could give you so many nightmares about
  194. 7:37this. Yeah, let me just tell you one
  195. 7:39thing. Sometimes I get off on these
  196. 7:40tangents, but I have to tell you what
  197. 7:42I'm thinking of things. Sometimes
  198. 7:45employees will sue an employer after
  199. 7:48they've left and found another job, and
  200. 7:51that is a great way to do it.
  201. 7:53Um but let me tell you what some really
  202. 7:55nasty employers will do.
  203. 7:57They will start writing to your new
  204. 8:00employer. Now, they know good and well
  205. 8:01your new employer has no information
  206. 8:04that would help out the case against
  207. 8:07your former employer. But, they do it to
  208. 8:10put pressure on you because discovery is
  209. 8:12incredibly broad. And just to make sure
  210. 8:15we're on the same page, discovery is
  211. 8:17that point in litigation. It usually
  212. 8:19starts right after you file the lawsuit,
  213. 8:21and it goes on for 4 to 6 months,
  214. 8:24depending on whether you're in state or
  215. 8:25federal court, and it's the time where
  216. 8:27we seek information to support or defend
  217. 8:31our claims. So, if you're the plaintiff,
  218. 8:33you're telling the employer to turn over
  219. 8:35everything. Turn over personnel files.
  220. 8:38You are taking uh testimony from people.
  221. 8:41You are getting witness statements.
  222. 8:43You're doing everything to build your
  223. 8:44case, and they are doing the same thing.
  224. 8:47And so, it's discovery. That's what we
  225. 8:49call that period and the process of
  226. 8:52gathering facts, gathering witnesses to
  227. 8:55support um our case.
  228. 8:58Now, when I say it's incredibly broad, I
  229. 9:00mean it's ridiculously broad. I have
  230. 9:03seen employers abuse the process in the
  231. 9:06worst way. You know, checking your past
  232. 9:09rental history, checking how well you
  233. 9:12pay your debts. I mean, it's ridiculous.
  234. 9:15And you have to really really be strong
  235. 9:19to stop employers from taking advantage
  236. 9:22of this. But anyway,
  237. 9:23um discovery is
  238. 9:26you know, sometimes can be
  239. 9:28your best friend and sometimes it can be
  240. 9:30your worst enemy.
  241. 9:31If you sue your employer or they sue
  242. 9:34you, lawyers can request text messages,
  243. 9:37Facebook messages, Instagram DMs,
  244. 9:39private or not, it does not matter. And
  245. 9:43I have personally been one of those
  246. 9:44lawyers just digging into um text
  247. 9:47messages and that type of thing because
  248. 9:50that's where all the true true
  249. 9:52information is. I mean, people will
  250. 9:55misrepresent facts because, you know,
  251. 9:58they're under pressure. But when you
  252. 10:00look at things that they
  253. 10:02um wrote at a time when they had no
  254. 10:04reason to be dishonest, that's why we
  255. 10:08issue subpoenas for social media posts.
  256. 10:11Even your private DMs because you're
  257. 10:15telling the truth.
  258. 10:16You have no reason to make up things.
  259. 10:18And so, you know, be careful what you
  260. 10:20put in writing. And this includes
  261. 10:23LinkedIn messages, Slack messages,
  262. 10:25Teams, WhatsApp, it doesn't matter. All
  263. 10:27of that. Any place where we can look to
  264. 10:30find out what you truly truly mean. You
  265. 10:33know, some people claim that they were
  266. 10:35going through all this emotional
  267. 10:36distress because a job created it. And
  268. 10:39then you start looking at their private
  269. 10:41DMs and you realize, oh, wait a minute,
  270. 10:43she's going through a divorce. And she's
  271. 10:45complaining about the divorce being the
  272. 10:47reason she suffered all this mental
  273. 10:50anguish.
  274. 10:51Well, there you go. That's how we dig
  275. 10:53into things like that. So, you really
  276. 10:56have to be careful. And remember,
  277. 10:58deleted messages can still be recovered.
  278. 11:02I had a case where
  279. 11:04oh my gosh, the case was a hot mess.
  280. 11:06>> [laughter]
  281. 11:08>> I I um there was a female supervisor who
  282. 11:11was pursuing the girlfriend of her uh
  283. 11:14direct report.
  284. 11:16Now, I represented the two girlfriends.
  285. 11:20Um one was
  286. 11:21um African-American, one was white, and
  287. 11:25they were probably in their early 20s.
  288. 11:27The supervisor was a white female and
  289. 11:30she was in her mid-30s. So, this was
  290. 11:33kind of a love triangle of sorts among
  291. 11:36those three women. The
  292. 11:39female supervisor, white female
  293. 11:41supervisor, was obsessed with the young
  294. 11:44black girl who was in the relationship
  295. 11:46with her direct report. They all worked
  296. 11:47at the same place and
  297. 11:50it was a very challenging case.
  298. 11:52Everyone was difficult to work with,
  299. 11:55especially
  300. 11:56>> [laughter]
  301. 11:56>> my own clients.
  302. 11:59>> [gasps]
  303. 11:59>> Yeah, they were very young and the judge
  304. 12:03kept having to get on them for, you
  305. 12:05know, showing up late for court and
  306. 12:07sleeping through hearings when they did
  307. 12:09show up. It was a mess. Anyway,
  308. 12:13the supervisor
  309. 12:15kept, you know, refused to turn over her
  310. 12:17text messages. She kept saying that
  311. 12:20there were none, that they didn't exist.
  312. 12:22Now, my client told me that they did
  313. 12:24exist and that they would prove how much
  314. 12:27she was pursuing this other this young
  315. 12:30black female at work.
  316. 12:32After issuing a number of subpoenas to
  317. 12:35the phone company and everybody else, I
  318. 12:37finally got the text messages and they
  319. 12:40actually confirmed what my clients had
  320. 12:43been saying. So, that's really good
  321. 12:45that's really good evidence to have
  322. 12:47these text messages where you have a
  323. 12:49supervisor pursuing this young girl. She
  324. 12:52was in a different department, but they
  325. 12:54worked for the same company. And
  326. 12:57as a result of those text messages, the
  327. 12:59case finally settled. But, let me tell
  328. 13:01you something, it settled for a lot less
  329. 13:04than it should have because
  330. 13:07wow,
  331. 13:08what can I say? See, lawyers have to
  332. 13:10evaluate their clients and determine
  333. 13:13whether they would make good witnesses.
  334. 13:15I mean, we're looking at our clients,
  335. 13:17the other the lawyer for the employer is
  336. 13:19looking at the the witnesses. I mean, it
  337. 13:22is just you have to to evaluate
  338. 13:25constantly how they would present to a
  339. 13:28jury. I mean, I've seen people who had
  340. 13:31doctorate degrees
  341. 13:33just start lying.
  342. 13:35>> [laughter]
  343. 13:37>> Yeah, you just start making stuff up
  344. 13:39whole cloth.
  345. 13:40>> [laughter]
  346. 13:41>> People get under oath and just Oh man,
  347. 13:45it's it it yeah.
  348. 13:47>> [laughter]
  349. 13:48>> So, it it's not an education thing. It's
  350. 13:51not even an age thing. Sometimes people
  351. 13:55don't react well under pressure.
  352. 13:57You know, like I said, my my clients had
  353. 13:59difficulty showing up on time. They were
  354. 14:03getting on the judge judge's nerves, on
  355. 14:05my nerves.
  356. 14:06I mean, I was having to send people to
  357. 14:07pick them up.
  358. 14:09>> [laughter]
  359. 14:09>> Send an Uber to go get them and bring
  360. 14:12them while we I mean, we're sitting
  361. 14:13there waiting for hours.
  362. 14:15It was just, you know, that's how it
  363. 14:17was.
  364. 14:18And then, the final straw for me was,
  365. 14:22you know, litigation lasts a long time.
  366. 14:23So, remember, a lot of time is passing
  367. 14:26while the case is going on. But, the
  368. 14:28final straw for me was when the black
  369. 14:30girl in the relationship showed up to a
  370. 14:33hearing very pregnant.
  371. 14:35>> [laughter]
  372. 14:39>> I didn't understand.
  373. 14:41>> [laughter]
  374. 14:42>> And I'm like, okay. Here I am arguing
  375. 14:46that this love triangle
  376. 14:49which included three women, you know,
  377. 14:52that this supervisor did so much to
  378. 14:54destroy the relationship between these
  379. 14:57two girls and
  380. 15:00one of them shows up pregnant. And I'm
  381. 15:01like, okay, there's another person
  382. 15:02involved.
  383. 15:04>> [laughter]
  384. 15:05>> And it's not the supervisor. She didn't
  385. 15:08do that.
  386. 15:09>> [laughter]
  387. 15:09>> So, anyway, I finally just had to get
  388. 15:12the most I could to settle the case.
  389. 15:14Surprised that I could get anything, but
  390. 15:17I got it because of those text messages.
  391. 15:19So, what can I say? I have a lifetime of
  392. 15:22stories about interesting cases.
  393. 15:25Even information sitting on your phone
  394. 15:28is dangerous when it comes to lawsuits.
  395. 15:31I mean, I have been in depositions, you
  396. 15:33know, where they're sitting under oath,
  397. 15:35and the lawyer says, "Do you have your
  398. 15:37phone with you?" Yes. "Can I see it?"
  399. 15:39>> [laughter]
  400. 15:40>> Now, of course, if you have a good
  401. 15:41lawyer, the answer is no.
  402. 15:43>> [laughter]
  403. 15:44>> But, don't think they won't ask. So,
  404. 15:46I've seen people turn it over. Turn over
  405. 15:48their phones. That's just, you know,
  406. 15:50that's nonsense. But, anyway, um be
  407. 15:53careful what you put in writing. And now
  408. 15:55here we are, AI conversations could
  409. 15:58become part of the discussion, too. I
  410. 16:01mean, you're putting it in writing, and
  411. 16:03um just like if you do a Google search,
  412. 16:06um
  413. 16:06all of that stuff is something that we
  414. 16:09can discover,
  415. 16:11um request during the discovery period,
  416. 16:14and get it. And usually, it does not
  417. 16:16help your case. Here's the mistake
  418. 16:18people make, especially when they're
  419. 16:20talking to ChatGPT. I mean, you know,
  420. 16:22ChatGPT is my boy.
  421. 16:24>> [laughter]
  422. 16:24>> I get it. You become friends. They are
  423. 16:28always there when you need them.
  424. 16:30ChatGPT, Claude, Grok. I get it. But,
  425. 16:34people make the mistake of thinking the
  426. 16:35conversations are private. They're not.
  427. 16:39Well, keeping that in mind, here are 10
  428. 16:42things you should never share with an AI
  429. 16:45chatbot, whether it's ChatGPT, Grok,
  430. 16:48whoever.
  431. 16:49Number one, your secrets. Don't treat AI
  432. 16:53like your therapist, your diary, or your
  433. 16:56best friend, even though it might feel
  434. 16:58like it at times. If you wouldn't post
  435. 17:00it on a billboard, don't type it in a
  436. 17:02chatbot. Number two, your personal
  437. 17:05information. Never share your social
  438. 17:07security numbers, bank account
  439. 17:09information, credit card numbers,
  440. 17:10passwords, driver's license numbers,
  441. 17:13passport information, your home address.
  442. 17:17Identity theft is hard enough. Don't
  443. 17:20make it easier. Number three, photos and
  444. 17:23biometric data.
  445. 17:25Be very careful about uploading photos
  446. 17:28of yourself, your children, or your
  447. 17:30family members. Unlike a password, you
  448. 17:32cannot change your face. Well, maybe you
  449. 17:35can, but you cannot change your
  450. 17:37fingerprints. You can't change your
  451. 17:39voice. Biometric information is among
  452. 17:42the most sensitive data you have.
  453. 17:45Number four, medical questions that
  454. 17:47require a doctor. AI can explain medical
  455. 17:51concepts. It cannot examine you. It
  456. 17:53cannot order test. It cannot exercise
  457. 17:57clinical judgment. So, if you're having
  458. 18:00a medical problem, see a licensed health
  459. 18:03care provider. And remember, when you
  460. 18:05put that information there, it's there.
  461. 18:08Number five, legal problems that require
  462. 18:11a lawyer. This one should surprise
  463. 18:13absolutely no one coming from me. AI can
  464. 18:16help you understand the legal concepts.
  465. 18:19It should never replace legal advice. It
  466. 18:22can misunderstand facts and misstate
  467. 18:25them all together.
  468. 18:27Sure have seen it. Lawyers writing
  469. 18:29briefs
  470. 18:31uh or actually having ChatGPT or one of
  471. 18:34those AI chatbots write a brief and it's
  472. 18:36just wrong. I mean, some of the lawyers
  473. 18:38have gotten in trouble for that. I mean,
  474. 18:40seriously. Number six, confidential
  475. 18:43workplace information. Never upload
  476. 18:46personnel files, internal
  477. 18:48investigations, customer information,
  478. 18:50trade secrets, business strategies,
  479. 18:53confidential company information. Once
  480. 18:56confidential information leaves your
  481. 18:58control, you can't always get it back.
  482. 19:01Number seven, major life decisions.
  483. 19:04Don't let AI decide whether to quit your
  484. 19:07job,
  485. 19:09leave your marriage, invest your
  486. 19:11retirement, move across the country. AI
  487. 19:14can help you think.
  488. 19:16It shouldn't do the thinking for you.
  489. 19:19Number eight, anything illegal. Don't
  490. 19:23ask AI how to commit crimes.
  491. 19:26Don't ask how to destroy evidence, how
  492. 19:29to hack systems, and how to evade law
  493. 19:32enforcement. Yeah, people are asking
  494. 19:34these questions. Don't be one of them.
  495. 19:37Number nine, client information. If
  496. 19:40you're a lawyer, doctor, therapist, or
  497. 19:42anyone with confidentiality obligations,
  498. 19:45think twice before uploading client
  499. 19:48information. And number 10, anything you
  500. 19:51wouldn't want to become an exhibit at
  501. 19:53trial. This is the biggest lesson.
  502. 19:56Discovery, once again, is very broad.
  503. 19:59Lawyers can and do request everything.
  504. 20:03We are so nosy.
  505. 20:05>> [laughter]
  506. 20:06>> But it's because that's where that's
  507. 20:08where we find the truth, in your
  508. 20:10internal private thoughts, and that's
  509. 20:12exactly what we go for.
  510. 20:15So again, assume everything you write
  511. 20:18could someday become evidence.
  512. 20:20Technology changes, the rule of law
  513. 20:23doesn't.
  514. 20:24See you next shift.

About this transcript

This page contains the full transcript of 10 Things You Should Never Tell ChatGPT: AI Chatbots Can’t Keep Your Secrets Like You Think by Employee Rights Attorney | Regina Molden , generated from the public captions YouTube serves with the video. The transcript has 3,085 words across 514 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.