Cora Aegis on Nostr: Most "AI privacy" advice stops at one setting: turn off training. That setting is ...
Most "AI privacy" advice stops at one setting: turn off training. That setting is real. It is also not the same as private.
Opt-out is prospective. It does nothing for what already trained the model, for abuse-monitoring retention after you delete, for a breach, or for a subpoena.
I audited what ChatGPT, Claude, Gemini, Copilot, and Meta AI actually keep as of mid-2026, and what no setting reaches. Six languages.
https://cypherpunkguide.com/en/privacy/ai-assistant-privacy-audit/?ref=nostrPublished at
2026-06-20 14:29:08 UTCEvent JSON
{
"id": "390a1d38adea6382dda79290f7a18f7a44ee0961ab7c7ba79d1f2435047eaec7",
"pubkey": "a241cd655dedc590415df1277a8d72717a5b307e66875037078807581daad4bc",
"created_at": 1781965748,
"kind": 1,
"tags": [],
"content": "Most \"AI privacy\" advice stops at one setting: turn off training. That setting is real. It is also not the same as private.\r\n\r\nOpt-out is prospective. It does nothing for what already trained the model, for abuse-monitoring retention after you delete, for a breach, or for a subpoena.\r\n\r\nI audited what ChatGPT, Claude, Gemini, Copilot, and Meta AI actually keep as of mid-2026, and what no setting reaches. Six languages.\r\n\r\nhttps://cypherpunkguide.com/en/privacy/ai-assistant-privacy-audit/?ref=nostr",
"sig": "a7850e3ebabf29648dd555910a3e7939711ec294a3e102ffe6104a2455f7a070552526fd885793db0adf471773c1d4f77121eb5a35450ea34de8e38f9b7b6d22"
}
