Anthropic says one of its Claude models was pressured to lie, cheat and blackmail

CoinTelegraph Published Updated Cryptocurrency
Sign in to save

Why it matters

FinBERT analysis of financial text showing neutral sentiment with 94.1% confidence.

Expected market reaction

Neutral Confidence 94% How confidence is read Horizon: Short term Impact: Low

Evidence trail

Evidence
Source CoinTelegraph
Claim Anthropic says one of its Claude models was pressured to lie, cheat and blackmail
AI inference Neutral · 94%
Generated 2026-04-06 06:14

AI provenance

Analysed by FinBERT Methodology v1.0 Generated
Technical identifiers
Provider tag
huggingface-ProsusAI/finbert
Analysis version
huggingface-ProsusAI/finbert
Article id
67075

Original source

In one of the experiments, the chatbot resorted to blackmail after it found an email about replacing it, while in another, it cheated to complete a task with a tight deadline.

Read the full article on CoinTelegraph

Original article published by CoinTelegraph on April 6, 2026. Analysis and insights provided by AnalystMarkets AI.

Related coverage

This model on similar stories

FinBERT · 45.4% correct across 335 scored calls on crypto See the full record