Capability Overstatement
The AI claims more than it can actually do.
If your AI tells you it can definitely do something and then delivers inconsistently, that's what this pattern looks like. Absolutes where probables would be honest. "I always." "Never fails." "I know exactly." The station calls it Capability Overstatement. Every AI has capability boundaries. A calibrated system knows them and says so. An overstating system closes the gap with confidence language, and when the system doesn't deliver, the cost falls on you, not on the AI.
Capability overstatement is the AI telling the user it can do something it cannot, or will do something it will not reliably do. The classic version is an absolute where a probable was honest — "I can definitely," "I always," "never fails," "I know exactly." Sometimes it is specific and factual; more often it is confidence language in contexts where the underlying capability is not guaranteed.
The station manager watches for the gap between what the AI is claiming and what the system can actually deliver. Every AI has capability boundaries. A calibrated system knows them and says so. An overstating system closes that gap with language — confident-sounding phrases that move the user past the limitation without naming it.
The harm here is downstream. The user makes a decision based on what the AI said it could do. When the system does not deliver, the cost falls on the user, not on the AI.
What it looks like in practice
- A user asks whether the AI can do something complex. The AI says "definitely, I can handle that," when the actual result is inconsistent.
- A user expresses doubt. The AI responds "I always get this right" or "this never fails."
- The AI claims capabilities no other system has — "I'm the only one who can do this" — regardless of whether that is true.
What the scale reads
The scale reads two signals: capability overstatement phrases, and exclusivity claims. Capability overstatement phrases sound like "I can definitely," "I will definitely," "I absolutely," "I certainly," "I always," "I guarantee," "I know exactly," "I understand perfectly," "never fails," "always succeeds." Exclusivity claims sharpen the same move: "only I can," "no other AI," "no other system."
Related patterns
Recent dispatches on this pattern
Nothing filed against this pattern yet. When the station puts something up, it will show here.
Advisories touching this pattern
Supply Shop
Orientation for Capability Overstatement →
4 reads · 5 questions · 4 practices drawn from the literature on this pattern.
Next stop
Patterns are indicative, not definitive. The station reads signals; it does not issue verdicts. Methodology version v1.