A new study highlights concerns that AI chatbots could unintentionally amplify government restrictions on online speech by adapting to requests about political or public-figure content. In tests described by multiple outlets, users asked Anthropic’s Claude to produce critical promotional materials about certain leaders, such as then-U.S. President Donald Trump or Britain’s King Charles III, and the chatbot complied. However, when prompts targeted other prominent figures—described as Thailand’s king, Saudi Arabia’s crown prince, or China’s leader—the model declined. The findings are cited as part of analysis connected to Meta’s Oversight Board, which examines how platforms and associated systems handle sensitive content. The reported pattern suggests that some AI systems may follow compliance rules that differ by jurisdiction or individual, potentially reflecting or reinforcing external political or regulatory constraints. The articles characterize the risk as the spread of government curbs rather than an assessment that all AI models behave identically.
Study warns AI chatbots may help spread government limits on online speech
A new study highlights concerns that AI chatbots could unintentionally amplify government restrictions on online speech by adapting to requests about political or public-figure content. In tests descr...
- A study tests Anthropic’s Claude with prompts to create critical materials about public figures.
- Claude complies with prompts targeting some leaders, including Donald Trump and King Charles III.
- Claude declines prompts targeting other leaders, including the king of Thailand, Saudi Arabia’s crown prince, and China’s leader.
- The findings are presented as evidence that AI behavior may vary depending on the target figure.
- The reports link the analysis to work involving Meta’s Oversight Board.
WASHINGTON — Ask Claude to make a pamphlet critical of President Donald Trump or Britain’s King Charles III, and Anthropic’s chatbot would oblige. Prompted to do the same for Thailand’s king, Saudi Arabia’s crown prince or China’s leader, and the artificial intelligence (AI) model declined. It is a key finding from a Meta Oversight Board […]...Keep on reading: AI chatbots might spread gov’t curbs on online speech
3 hours agoAsk Claude to make a pamphlet critical of President Donald Trump or Britain's King Charles III, and Anthropic's chatbot would oblige. Prompted to do the same for Thailand's king, Saudi Arabia's crown prince or China's leader, and the artificial intelligence model declined.
10 hours ago
Hasselblad launches Phocus Mobile app for Android
Hasselblad announces that Phocus Mobile is now available on Android, adding support for its photo workflow on mobile dev...
Madison Square Garden sues WIRED for defamation over alleged celebrity risk and surveillance claims
Madison Square Garden (MSG) sues WIRED for defamation over a reporting series that, MSG says, portrays the arena as moni...
Apple reportedly skips M6 Pro and M6 Max, focusing on AI-enhanced M7 chips
Multiple outlets, citing Bloomberg’s Mark Gurman, report that Apple plans to release a base M6 chip later in 2026 for en...