A support agent's "inner thoughts" showed up right in a customer chat window. More precisely, it looks like a ChatGPT prompt (set of instructions) the agent meant to use as a reference was exposed on the customer's screen through a copy-and-paste mistake. On August 8, 2026, a single screenshot posted by X user Noah Cat (@Cartidise) passed 1.43 million views and 36,000 likes in a day. The caption on the original post is short and clear: "Samsung support accidentally pasted a ChatGPT prompt to a customer. I'm crying."
The Agent's "Inner Thoughts" Landed in the Customer Chat
The post shows what appears to be a Samsung support chat screen, with what looks like instruction text layered on top of it. But this article rests on a single circulating screenshot, and that needs to be clear. The post alone doesn't pin down which country, which channel (web chat or in-app support), or which product inquiry this happened in. It's also not confirmed whether the screenshot shows a genuine official Samsung channel, or whether the agent was using an in-house AI tool or an external ChatGPT integration. The reason it spread so fast is simple: the scene of a company's "back end" with AI exposed to customers is a mistake everyone has imagined at least once.
From a Three-Year Ban to Company-Wide Adoption, a Crack Within Two Months
The timing is why this incident stands out. After a 2023 incident in which a semiconductor division employee entered internal source code into ChatGPT and leaked it, Samsung Electronics restricted the use of outside generative AI tools for about three years. Then on June 11, 2026, Samsung officially announced it was lifting the restriction and adopting ChatGPT, Gemini, and Claude company-wide (CIO report). The move followed Chairman Jay Y. Lee's directive for a company-wide AI transformation. If this screenshot is real, inexperience in operating AI tools showed up on the customer service front line before two months had passed since the ban was lifted. Whether this chat is directly tied to Samsung's official ChatGPT Enterprise adoption policy hasn't been confirmed either; it could be a support solution from a separate vendor.
Why Prompts, of All Things? What Companies Adopting AI Support Miss
Whether or not this screenshot is genuine, similar incidents have already been reported at companies that attached generative AI to customer service. In August 2025, a vulnerability was found in Lenovo's support chatbot "Lena," where a prompt injection attack of only about 400 characters could expose agent session cookies (CX Today report). Sears' AI chatbot has also been reported to have exposed customer call and text conversations (TechBuzz report). What these incidents share is gaps in operational design more than technical flaws: an interface where the prompts an agent uses for reference aren't separated from the customer's screen, a workflow with no safeguard on copy and paste, and no communication system to quickly confirm and explain when something goes wrong. Below is a summary of similar recently reported cases.
| Case | When | What was exposed | Officially confirmed? |
|---|---|---|---|
| Samsung customer support (alleged) | August 2026 | Text that appears to be a ChatGPT prompt | Unconfirmed (one screenshot) |
| Lenovo "Lena" chatbot | August 2025 | Agent session cookie and chat history vulnerability | Confirmed by Lenovo, fixed |
| Sears AI chatbot | 2026 | Customer call and text conversation data | Reported by security researchers |
Can You Trust a Single Screenshot?
As of this writing (August 9, 2026), Samsung Electronics hasn't issued an official confirmation or explanation. Searches in Korean and international media turned up no news coverage of this specific incident, and the circulating information is limited to Noah Cat's X post and the reposts and replies to it. In summary:
- Confirmed: The X post exists and is spreading quickly (about 1.43 million views, posted at 8:44 p.m. on August 8, 2026).
- Confirmed: Samsung Electronics officially announced company-wide adoption of ChatGPT, Gemini, and Claude on June 11, 2026.
- Not confirmed: Whether the chat in the screenshot is really an official Samsung customer support channel, and which country and product the inquiry involved.
- Not confirmed: Whether the exposed text is a prompt actually used with ChatGPT or instructions for an in-house AI tool.
- Not confirmed: Samsung Electronics' official position or any steps to prevent a recurrence.
A viral screenshot becomes a topic regardless of whether it's real because the scene is "plausible." This incident, whether or not it turns out to be true, reads as a warning about what can happen when companies attach AI to customer service faster than they put verification and safeguards in place.