Anthropic’s Sarah Heck highlights AI safety as key voter concern ahead of midterms

Anthropic isn’t just building AI models. It’s trying to shape the political conversation around them. Sarah Heck, the company’s Head of Policy, has been making the case that AI safety isn’t some niche tech concern but a legitimate voter priority heading into the 2026 US midterm elections. The numbers behind the messaging Anthropic’s commitment goes well beyond press statements. The company has pledged more than $40 million to Public First Action, a bipartisan organization focused on advancing AI education and safety measures. Anthropic employees have also been opening their personal wallets. By the end of Q1 2026, staffers had made over 302 individual political donations totaling more than $880,000, with the average contribution exceeding $12,500 per donor. Those contributions have flowed toward candidates aligned with AI safety initiatives. On the product side, Anthropic has been tuning its Claude models to handle election-related queries with unusual precision. An April 2026 update showed Claude Opus 4.7 responding correctly to 100% of legitimate election prompts, while Sonnet 4.6 hit 99.8%. The models also incorporate web-search functionality for candidate information, achieving...