Avoid letting AI access texts; use local models or ZDR providers instead
uwukko · x · 2026-08-24
A user warns against granting AI tools access to personal data like texts, especially given vague privacy policies. Citing ProtonPrivacy, the post emphasizes not letting ChatGPT read messages. It suggests that for AI agents handling messages, one should use open-source tools with local models or inference providers with verified Zero Data Retention (ZDR).
More from Safety
- Copilot Exfiltration via CSP Bypass and Memory Poisoning — wunderwuzzi23 · 2026-08-24
- Artificial Intelligence and Its Effects on Employment - France Government — 233C · 2026-08-24
- What tasks should AI agents never be allowed to do autonomously? — omnidimension85 · 2026-08-24
- Teachers targeted by students using AI to generate fake nudes: Report — RebeccaBellan · 2026-08-24
- LLMs Are Now Attack Surface: Why the OWASP Top 10 for LLMs Matters — mclynd · 2026-08-24
- Decentralizing AI Identity: Why Nostr is the Bedrock for Agent Continuity — RileyRalmuto · 2026-08-24