OpenAI support's fake 'human escalation' asks for screenshots three times
dynamitecGbR · reddit · 2026-09-27
A developer complained about OpenAI support: after explicitly requesting human escalation, the support bot assured him it would comply, then sent replies clearly written by AI with no disclosure.
- His issue was a Codex tooling bug: closeagent is documented as available but absent from the visible tool list — not something a screenshot can show.
- The support AI asked three times for a screenshot before forwarding to "another human support member," and ignored links to Codex repo bug reports describing the same problem.
- The author likened the experience to talking to Alexa: the promised human never appeared.
More from Companies & People
- Yale AI course demand doubles classroom twice in one week, materials to go public — aminkarbasi · 2026-09-27
- ShellHacks draws ~2K student builders in Miami — but 9K-student FIU still can't keep CS grads local — KyeGomezB · 2026-09-27
- No Priors: who a company celebrates — investors vs engineers — decides who it attracts — No Priors · 2026-09-27
- OpenAI's DevDay agent rumored to skip $20 Plus tier, starting at $1,200/year — FamilyNP · 2026-09-27
- All-in Podcast: Dario 'sabotaging' Anthropic IPO, odds fall 21 points in 3 weeks — trevposts · 2026-09-27
- ThePrimeagen denies taking money from dhh's Omarchy: "$0.00 total" — AccBalanced · 2026-09-27