Kurtis The Quant

Nah, it’s not his fault at all. Bro was just trying to do a solid for his father in law by wiping an old iPad that belonged to a deceased 90 year old man, clearing the personal data, and possibly putting the device to good use. That is normal family and estate cleanup, not some criminal mastermind plot. Activation Lock exists specifically to prevent theft, so the moment he asked ChatGPT how to remove the lock when the owner was dead and there was no receipt or will, the automated filters went full paranoid mode and slapped him with a “Cyber Abuse” label. OpenAI’s classifiers can be notoriously trigger happy when it comes to anything involving device unlocking, reverse engineering, or cybersecurity related discussions. There are reports of people doing completely legitimate things, such as coding their own projects, preparing cybersecurity job applications, or working on hardware they own, receiving similar warnings or worse, sometimes followed by appeals being rejected within an hour through what appear to be template responses. The system saw keywords and context that suggested possible unauthorized access, ignored the backstory about the deceased owner and the informal agreement, and defaulted to caution. He even explained the situation in the chat, but nuance clearly is not the classifier’s strongest suit right now. Practically speaking, he has already appealed and lost, so pursuing further human review through the usual channels may be a dead end. The real solution for the iPad is still Apple Support, which can explain whether the device can be unlocked with appropriate documentation. Proof of death and evidence of the relationship may help, although an informal agreement alone does not guarantee an exception. As for his OpenAI account, the safest move is to respect the warning and avoid further prompts that could be interpreted as attempts to bypass device security. People who believe they have encountered a false positive can also use the available official support or appeal channels to seek clarification. The underlying problem he pointed out is real. AI safety systems can struggle to distinguish between someone helping a family member deal with a deceased person’s old tablet and someone attempting to gain unauthorized access to a device. Legitimate requests deserve careful contextual assessment, not just keyword matching. Safety matters, but so does getting the context right. If the system gets that balance wrong, legitimate users end up paying the price.