YouTube2Text

Claude Code E-postamı İki Siteye Göndermiş — Transcript

by Selma Kocabıyık · 1,425 words · 218 segments · language en · Watch on YouTube

Full transcript

  1. 0:00You know that on this channel we
  2. 0:01generally talk about agents, or I show
  3. 0:03small agent automations and explain how
  4. 0:04you can build them using artificial
  5. 0:06intelligence. So, I think what I'm going
  6. 0:08to talk about in this video interests
  7. 0:10both you and me very much. In the last 3
  8. 0:13days, two major security issues
  9. 0:14regarding Claude have emerged.
  10. 0:16Now, we will discuss two of these today
  11. 0:18; the first is that Claude has been
  12. 0:20sending your and my email addresses to
  13. 0:22certain sites you visit, and the second
  14. 0:24is that when working with Claude, it
  15. 0:26sometimes asks you a question, like "
  16. 0:28should I do this" or "should I do that,
  17. 0:30" and you say "yes, don't ask again."
  18. 0:34Well, when we do that, your information
  19. 0:36, along with your passwords, is written
  20. 0:39into a file, and that file is sent to
  21. 0:41over a thousand GitHub projects.
  22. 0:45Yes, hearing this, the first thing I did
  23. 0:47was check my computer, and yes, my
  24. 0:48Claude, thank you very much, had indeed
  25. 0:50sent my information to other websites.
  26. 0:53So, how did this happen? I'll explain it
  27. 0:55right away. On November 5th, a developer
  28. 0:57posted a complaint on Claude's GitHub
  29. 0:59page.
  30. 1:01It said something like, "I'm receiving
  31. 1:03dozens of scam emails every day because
  32. 1:05Claude gave my email to websites without
  33. 1:07my permission."
  34. 1:09Now, I don't know how true or false this
  35. 1:11is, but since the same thing happened to
  36. 1:13me, I think it's true. Now let me
  37. 1:15explain how this happens logically. When
  38. 1:18you visit a site, you send a small ID
  39. 1:20card to that site. Think of it like an
  40. 1:23ID card that says "I am this program,"
  41. 1:25and this is called a user agent. Now,
  42. 1:27Claude naturally knows which email you
  43. 1:29used to log in. For example, you use
  44. 1:31Claude to do research. Am I on fire? You
  45. 1:34send your agent to the internet and tell
  46. 1:35it, "Do research on this topic for me."
  47. 1:38So what happens? When Claude enters that
  48. 1:40website, it writes your email address
  49. 1:42into that ID card.
  50. 1:45Of course, neither you nor I go to
  51. 1:47Claude and give instructions like "go to
  52. 1:48any website you want and put my email
  53. 1:50there," you don't authorize this either
  54. 1:52.
  55. 1:54But whether due to permissions we gave
  56. 1:55previously, or because the agent thinks
  57. 1:57it doesn't need authorization at that
  58. 1:59point, it can give out your email like
  59. 2:01this.
  60. 2:02Now, normally Anthropic added a check
  61. 2:04for this. It is supposed to ask you
  62. 2:06before it sends your email. But the
  63. 2:08person who complained was already using
  64. 2:10the version that had this check, and
  65. 2:12they experienced the same problem anyway
  66. 2:14.
  67. 2:14So, the issue is not completely closed
  68. 2:16then. Claude saves all conversations on
  69. 2:18the computer.
  70. 2:20Naturally, we—or at least my agent and I
  71. 2:22—scanned the entire computer, and in a
  72. 2:24possible scenario, we looked at past
  73. 2:26conversations and the underlying
  74. 2:28codebase to see if the agent had really
  75. 2:30sent my email to a site.
  76. 2:33The first of these was on October 26th;
  77. 2:35while preparing a video, Claude went to
  78. 2:37Wikipedia to search for images and left
  79. 2:39my email there.
  80. 2:41The second one happened just today.
  81. 2:44While researching the video we were
  82. 2:46going to make this week, it placed my
  83. 2:47email address on the SEC website.
  84. 2:50To be fair, those sites were actually
  85. 2:52asking for contact information anyway.
  86. 2:55Cloud knows the rules to complete the
  87. 2:56task and executes the instructions. But
  88. 2:59it does this without asking me, and who
  89. 3:01knows what else it’s doing in the
  90. 3:03background without my consent; I should
  91. 3:05have been the one to decide which site
  92. 3:07my data goes to, and this is just the
  93. 3:09part we see. So that was the first part
  94. 3:11. Let’s move right on to our second part
  95. 3:13. The second part is more problematic,
  96. 3:14by the way. When Cloud Code wants to run
  97. 3:16a command, it normally asks you. Now,
  98. 3:18the first option here is "Yes, don't ask
  99. 3:21again." Most of us click this because we
  100. 3:23don't want to have to give approval
  101. 3:24every single time. Some of us even go as
  102. 3:25far as granting permission for
  103. 3:26absolutely everything. Okay, that's not
  104. 3:28a problem. If you don't have any
  105. 3:29important data, it's actually not an
  106. 3:30issue.
  107. 3:31But if you do have important data,
  108. 3:32that's when we run into trouble.
  109. 3:35There is a file called [settings.json].
  110. 3:36This file contains all the rules within
  111. 3:39that project and is a file that
  112. 3:40decisively affects what the agent can
  113. 3:42and cannot do.
  114. 3:44The agent goes and writes into that file
  115. 3:46any password or API key contained in a
  116. 3:49command so it can use it whenever it
  117. 3:51needs to.
  118. 3:52By the way, this is a file on your local
  119. 3:54machine, and it is not an encrypted file
  120. 3:56. So, in the event of a possible virus
  121. 3:58attack or any other unauthorized process
  122. 4:01, anyone—an outsider—can reach and find
  123. 4:04all your information in this
  124. 4:06settings.json file.
  125. 4:08They can gain access to all your API
  126. 4:09keys and all your passwords from here.
  127. 4:13On November 5th, a security firm scanned
  128. 4:15104,000 such files that were public on
  129. 4:17GitHub, and 1,102 of these files
  130. 4:19actually contained a real password or
  131. 4:22key.
  132. 4:23Look, I'm talking about people's
  133. 4:24personal information here.
  134. 4:26With Git history included, a total of
  135. 4:28over 2,500 secrets—meaning your data and
  136. 4:30passwords—were found.
  137. 4:32Now, the most interesting detail is this
  138. 4:34: two popular password scanners on the
  139. 4:36market failed to catch one-fifth of
  140. 4:37these files at all.
  141. 4:39Naturally, I immediately scanned my own
  142. 4:41computer after this news, and 18 of
  143. 4:43these files turned up.
  144. 4:45Now, there are exactly 179 "don't ask
  145. 4:47again" commands in my main folder, and I
  146. 4:49don't even remember them. For me, the
  147. 4:51good news at least is that none of them
  148. 4:53went to GitHub. It seems there was a
  149. 4:55general rule on my computer that hides
  150. 4:57this file in every project. I don't know
  151. 4:59if it comes from Apple or from my own
  152. 5:01excessive meticulousness, but at this
  153. 5:03point, nothing happened to my passwords
  154. 5:05.
  155. 5:06By the way, for my scenario, I've
  156. 5:07prepared a separate file explaining how
  157. 5:09you can check this if you want to verify
  158. 5:11it yourself.
  159. 5:12You can find that in the description
  160. 5:13section as well. Definitely try it on
  161. 5:15your computer and run a scan. A quick
  162. 5:17scan doesn't hurt. Now, let’s wrap this
  163. 5:19up quickly. I will explain what you can
  164. 5:21do in 4 steps. First, the first thing
  165. 5:23you need to do is check your version. It
  166. 5:25is important that both your computer is
  167. 5:27up to date and your Cloud Code—or it
  168. 5:28could be Codex, by the way—is updated.
  169. 5:30We are always checking based on Cloud
  170. 5:32Code, but whatever agent application you
  171. 5:34are using, make sure it is up to date.
  172. 5:36Moving on to our second step, check if
  173. 5:37this file has been pushed to GitHub in
  174. 5:39your project.
  175. 5:40Third, open the file, and if you see any
  176. 5:42passwords or key information, it doesn't
  177. 5:45matter, delete them, change them, or set
  178. 5:47them up differently.
  179. 5:48By the way, just deleting the file isn't
  180. 5:50enough. You must also remove it from
  181. 5:52your Git history.
  182. 5:54Likewise, if you are writing code for a
  183. 5:56company or a project, you need to pay
  184. 5:58extra attention to such things.
  185. 6:01Moving on to the fourth point: do not
  186. 6:03write passwords inside commands.
  187. 6:05Passwords should only reside in an
  188. 6:07environment file.
  189. 6:08No matter what, do not type any of your
  190. 6:10passwords into the chat or within a
  191. 6:11command. Sometimes I do this. I give the
  192. 6:14agent a key or a password, for instance
  193. 6:15, to write into the environment file.
  194. 6:18Afterwards, I make sure to delete that
  195. 6:20part of the chat where I wrote it.
  196. 6:22The agent makes sure, and I make sure
  197. 6:23too. We delete that part permanently. So
  198. 6:25, be very careful about where and how
  199. 6:27you save your passwords. Now, when
  200. 6:29agents start doing work for us, in a way
  201. 6:31, they also start speaking on our behalf
  202. 6:34.
  203. 6:34Therefore, they might share our email,
  204. 6:36give out our password, or send files
  205. 6:38externally.
  206. 6:40Because for them, it's not about being
  207. 6:41malicious, but rather seeing these as
  208. 6:43steps that must be taken to complete the
  209. 6:45task.
  210. 6:46That is why you need to be very aware of
  211. 6:48what you provide to your agent and keep
  212. 6:50your information extra secure.
  213. 6:52I think it would be great if we paid a
  214. 6:53little attention to this issue and
  215. 6:55dedicated some of our time to it. Please
  216. 6:57share your thoughts on the subject in
  217. 6:58the comments, and I will see you in the
  218. 7:00next video.

About this transcript

This page contains the full transcript of Claude Code E-postamı İki Siteye Göndermiş by Selma Kocabıyık, generated from the public captions YouTube serves with the video. The transcript has 1,425 words across 218 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.