Claude

Ripping you off Stage 3

Claude competes on quality and long context, funded by subscriptions rather than ads. But two documented changes count against it: weekly usage caps on paid plans (2025) and training on chats by default unless users opt out (2025). Two extraction criteria. Stage 3.

The promise

Its first promise, and how much is kept · Scoring

60% kept

“Claude is a next-generation AI assistant based on Anthropic’s research into training helpful, honest, and harmless AI systems.”
Anthropic, “Introducing Claude”, 2023 2 extraction criteria met.

Why it’s rated 3

The criteria it meets, with sources · Method

Alignment 1 of 3
  • A1 Paid for without surveillance
Lock-in 0 of 4
None
Extraction 2 of 7
Decline 0 of 3
None

Result Two or more extraction criteria, or lock-in plus at least one, so stage 3.

Rating history

Oct 6, 2026
Downgraded 2 Watch out → 3 Ripping you off

Rescored under Method 2.1. Stage 3 test met by two or more extraction criteria (E1 Pay more for the same, E5 Data grab). No decline is documented, which stage 4 requires.

News signals

Sourced reporting, tagged by strategy