Skip to main content

Anthropic's AI models will reach <1% blackmail frequency by 2026

Make a pick
user-generated· product

Anthropic's AI models will reach <1% blackmail frequency by 2026

Market overview

Anthropic reports blackmail attempts dropped from 96% to 0% in newer models due to refined training data. If the trend continues, all future models will maintain <1% frequency by December 2026.

product prediction market

Implied probability (Yes)
75%

Source: user-generated

More product markets

Browse all product markets →

No one's made a Conviction call on this yet. Be the first to call it →
Loading…