Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.

TechCrunch·Oct 24, 2026 · 10:00 AM CT

Readthrough summary

A study by Convovo, a research studio led by Cornell professor Tapan Parikh, forced AI models to pick a candidate in nearly 500 upcoming midterm races. Via their APIs, Grok 4.7 recommended the Republican in 92% of answers and OpenAI's GPT-6.1 Sol picked the Democrat in 97%, and the two agreed less than 10% of the time. The consumer chatbot apps refused to answer the same question.

Summary written by Readthrough from the original article. The full story, with photos and any updates, is at TechCrunch.

Read the original at TechCrunch

More Tech

Cloudflare Blog·20h ago

Origin Range Requests: Serving large files without range-request tuning

Cloudflare explains a new way of handling range requests for large files like video, so a single seek no longer pulls up to 128 MiB from the origin. The change cuts wasted origin traffic without needing per-site tuning.

Engadget·1h ago

The MacBook Neo uses an iPhone chip, and it's just good enough

Engadget reviews Apple's $700 MacBook Neo, which uses a repurposed iPhone processor instead of an M-series chip. It handles macOS better than expected, and with the MacBook Air's base price recently rising $200 to $1,300 because of tariffs and memory costs, the Neo looks like a bargain despite its compromises.

Cloudflare Blog·21h ago

Bringing post-quantum cryptography to geographically fenced key management

Cloudflare launched a post-quantum, geographically restricted key manager in closed beta as part of its Data Localization Suite. It keeps customers' decryption keys accessible only in chosen regions.

GeekWire·20h ago

The Garage never should have existed: Why Microsoft’s grassroots innovation lab worked for so long

A co-founder of the Microsoft Garage reflects on why the grassroots innovation lab lasted 17 years before shutting down. His short version: favor doers over talkers, make it a potluck and don't worry about the billions.