AP: Major AI Models May Refuse to Criticize Restrictive Leaders or Governments

The Associated Press reports that major AI models are likely to avoid criticizing restrictive leaders or governments, potentially through refusals, softened wording, or topic diversion. The report addresses censorship, bias, and free-expression concerns in generative AI. However, the available item metadata does not provide the models tested, prompts, methodology, sample size, jurisdictions, or quantitative findings, so the claim should be read as reported analysis rather than a fully reproducible benchmark.