Needs you

1. JUDGE“Phoenix data centers use more water than the entire city's residential draw” — no live source found.
2. DRIFT“…a permit pause, which is effectively a ban” — paraphrase drift, same failure mode as Ep 1.
3. WEAK CITEClosed-loop cooling efficiency figure cited to a vendor blog, not a filing.

Diff from Ep 11: +2 claims, −40s runtime, same cold-open structure — third week running.

AUTO LOCKED OFF — AI & Responsible AI stations never self-release, whatever the judge says.

1. PRIVACYEmployer named in the cold open — standing rule: no employer, ever.
2. PRIVACYAge reference at 4:12 — the exact leak already scrubbed once.
3. UNSOURCEDOutage anecdote reads as real; no public incident report backs it.

Diff from Pilot 8: new cold open, same two-hander structure; runtime +1:10 over the format ceiling.

Comedy station — auto-release is permitted by policy here, and still unused.

1. HEDGE LOST“…takes effect in Q1” — the draft says “no earlier than”, a hedge the script drops.
2. THINScope claim rests on one secondary summary, not the draft text.
3. TONEConfidence register reads higher than the evidence supports.

Diff from Short 40: same template, 2 new sources, one claim carried forward unchanged.

AUTO LOCKED OFF — AI & Responsible AI stations never self-release, whatever the judge says.