The votes are in: AI will hurt elections and relationships

Key findings

It has been reported that Stanford University's Institute for Human‑Centered Artificial Intelligence (HAI) warned in its 2026 AI Index Report that AI adoption has raced ahead — 53 percent of the population in three years — while safety and governance lag. Documented AI incidents reportedly climbed to 362 in 2025, up from 233 in 2024, and the report bluntly states that "responsible AI is not keeping pace with AI capability." Ouch. Rapid adoption, rising harms — the math does not look friendly.

Where the pain lands

Experts and the public apparently disagree on almost everything about AI’s future, except two things: elections and personal relationships will suffer. It has been reported that 64 percent of Americans expect AI to reduce jobs over the next two decades (only 5 percent expect more), while experts are more split. Trust in government regulation is low — just 31 percent of US respondents said they trust their government to regulate AI responsibly. China is allegedly catching up to the USA in capability, too. Scary? You bet. But also unsurprising.

The technical wobble

Beneath the headlines, the tech story is messy. Models have gotten startlingly good at code — SWE‑bench scores rising toward 100 percent — but benchmarks show huge variation on honesty and uncertainty: hallucination rates ranged from 22 to 94 percent across models on the AA‑Omniscient Index. It has been reported that attorneys were even rebuked by the US Sixth Circuit for using AI to fabricate citations. And for a dose of humility: GPT‑5.4 High reads analog clocks correctly only about 50.6 percent of the time, humans around 90 percent; household robots succeed in barely 12 percent of tasks on BEHAVIOR‑1K. Can you trust systems that can’t tell time? I wouldn’t bet my ballot on them.

The bottom line

The 423‑page report — written by human researchers with help from ChatGPT and Claude and reportedly backed by industry funders — pulls no punches: adoption has outpaced safety, public anxiety is real, and the most concrete near‑term harms look political and personal. So what now? Regulators, researchers and companies have to close the gap. Otherwise the machines won’t just be buggy — they’ll be disruptive in the worst possible ways.

Sources: The Register