Anthropic published an update on April 24 detailing how Claude is prepared for the 2026 US midterm election cycle, including evaluation results and new guardrails. Opus 4.7 and Sonnet 4.6 scored 95% and 96% on political impartiality evaluations and responded appropriately to 100% and 99.8% of election-misuse test prompts; both models refused nearly every task in a new category of test that simulated autonomous influence operations — multi-step campaigns run without human prompting. Election-related queries now surface a TurboVote banner, and Anthropic is partnering with The Future of Free Speech and the Collective Intelligence Project on broader behavioral review.