Anthropic strengthens election safeguards ahead of 2026 US midterms

Anthropic has announced enhanced election safeguards ahead of the 2026 US midterms, emphasizing its commitment to preventing abuse and ensuring balanced political discourse. The company analyzed 250,000 conversations on its platform, revealing that 5.75% of US interactions involve politics, public policy, or social issues, with 0.51% specifically about elections. Political bias and misuse were rare, appearing in less than 0.01% of conversations. Anthropic now integrates official voting information from Democracy Works into its responses, providing state-specific details on polling places and ID requirements. It also displays real-time election results on election night. The company’s latest models, including Claude Opus 5.5, Sonnet 5.5, and Haiku 5.5, scored 99% on political evenhandedness tests. Independent evaluations by the Future of Free Speech, the Foundation for American Innovation, and the Collective Intelligence Project highlighted areas for improvement, such as inconsistent hedging and unsolicited counterarguments. Anthropic has updated its Usage Policy to clarify restrictions on deceptive practices and launched a guide for political campaigns. It also enhanced classifiers to detect multi-step election abuse and strengthened protections against foreign interference. The company reported stopping efforts by foreign actors to manipulate US midterm elections through AI-generated content.