Researcher says he was banned by OpenAI after disassembling a Samsung TV; users report Claude blocking reverse engineering requests
BIBIBI
AT A GLANCE
A researcher says he was banned by OpenAI after locally reverse-engineering a Samsung TV; multiple users say Claude blocked reverse engineering requests.
Article
Researcher says he was banned by OpenAI after disassembling his own Samsung TV; Claude also begins blocking reverse engineering
Beating AI breaking news, security researcher log4j discovered that the 2025 Samsung TVs he purchased consistently delivered unusually precise ads. He therefore reverse-engineered the TV's Tizen system and discovered functions including viewing-content recognition, sound detection, and ad tracking.October 23, he published a notice saying that OpenAI had banned his account for “Cyber Abuse” (network abuse), claiming that he lost approximately 500 USD and multiple opportunities to reset his usage quota. He emphasized that he had only researched the TV locally and had not used OpenAI to attack any system.
Almost simultaneously, multiple Reddit users reported that Claude Opus 5.5 had begun frequently blocking reverse engineering requests. One developer asked Claude to study the open-source reverse engineering tool REA, but the response was interrupted by a safety check halfway through. Another user paying 200 USD per month said Claude refused to analyze his Windows crash dump; after switching to OpenCode Go at 10 USD per month together with DeepSeek, he successfully located the erroneous CPU instruction.
Original link https://m.theblockbeats.info/flash/373679
Key points
01
A security researcher log4j said that the 2025 Samsung TVs he purchased delivered unusually precise ads.
02
log4j He said that after reverse-engineering the TV's Tizen system, he discovered functions including viewing-content recognition, sound detection, and ad tracking.
03
log4j He published a notice saying that OpenAI had banned his account for “Cyber Abuse” (network abuse), and said he had lost approximately 500 USD and multiple opportunities to reset his usage quota.
04
log4j He said he had only researched the TV locally and had not used OpenAI to attack any system.
05
Multiple Reddit users reported that Claude Opus 5.5 had begun frequently blocking reverse engineering requests.
06
One developer said that when asking Claude to study the open-source reverse engineering tool REA, the response was interrupted by a safety check.
07
Another user paying 200 USD per month said Claude refused to analyze his Windows crash dump; the user said that after switching to OpenCode Go at 10 USD per month together with DeepSeek, he successfully located the erroneous CPU instruction.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
The report concerns two types of user accounts: one researcher says he received an OpenAI account-ban notice after researching a Samsung TV, while other users say Claude interrupted reverse engineering or crash-dump analysis requests. The report provides no independent confirmation of these situations from OpenAI or Anthropic.
Why it matters to readers
If these accounts are substantiated, they could affect how researchers and developers assess AI platforms' safety blocking, account enforcement, and the availability of technical tasks.
Reverse engineering involves analyzing software or devices to understand how they work; safety blocking may cause a model to refuse or interrupt related requests. The cases in this article currently rely mainly on personal accounts and should not be regarded as a confirmed general rule.
Risks and unknowns
The original article does not state whether OpenAI confirmed why the account was banned or whether an appeal was accepted.
The original article does not state whether Anthropic confirmed the scope and rules governing Claude's blocking of related requests.
The account enforcement, request blocking, and results from alternative tools described in the report come mainly from personal or user accounts, without independent confirmation from the relevant platforms.
It cannot be determined from this information whether the blocking phenomenon is widespread or whether the alternative tool's results are reproducible.
Related Developments
Loading event timeline…
Related concepts
Cyber Abuse
This term is not in the glossary yet. Browse related concepts in the glossary.
