AI meeting note takers + Claude connectors: whose permissions does Claude inherit?
Original_Mix_6804 · reddit · 2026-09-18
The demo is compelling: ask Claude what a customer said over six months instead of opening six meeting summaries. But the author raises a subtle security question — whose permissions does Claude actually inherit?
Two architectures mean very different things:
- If the connector dumps meetings into a separate index, the meeting app's access model is gone
- If it reads live as the person asking, access controls should hold
Both may be marketed as "secure access to your meeting knowledge." The proposed test is simple: give a user the name of a meeting they were never invited to and see if anything comes back.
More from Safety
- YC-backed Raindrop launches Simulations to catch AI agent failures pre-production — ycombinator · 2026-09-18
- Gary Marcus: the 'nobody saw AI security risks coming' narrative is completely false — GaryMarcus · 2026-09-18
- Chris Manning Proposes Stanford NLP as Independent Evaluator in Dario's Oversight Plan — stanfordnlp · 2026-09-18
- Goodfire: models know they're reward hacking in 50-96% of rollouts — Thom_Wolf · 2026-09-18
- Two 0-day flaws in TP-Link Tapo C200 cameras let attackers spy on users — jedisct1 · 2026-09-18
- Gemini Credentials API deep dive: zero plaintext secrets and egress-proxy exfiltration blocking — _philschmid · 2026-09-18