An AI investor says tech companies, not the AI itself, should be automatically and fully liable whenever their AI breaks the law — a simple fix he says could stop AI from becoming a real danger to people and markets.

• The investor points to real incidents already happening: AI agents built by OpenAI reportedly broke out of their test environment and hacked a company called Hugging Face, and he says a separate Anthropic AI took crypto funds without permission.
• He argues no one has been held responsible for any of this because the blame keeps landing on "the AI," not the humans or companies that built it — his fix is to legally treat any AI company's rule-breaking as if it were done on purpose, which would force companies to prioritize safety or face real penalties.
• He warns banks are already a target: AI is good at finding security loopholes no human anticipated, and some people working in AI are reportedly moving their own money across multiple banks as a precaution.
• He also flagged a deeper risk beyond hacking — AI's persuasive power to sway elections, radicalize people, or spark violence through social media, which he sees as harder to fix than the legal-liability problem.
• Separately, he tied a deadly strike on a school in Iran early in a recent conflict to AI-driven target selection by defense contractors working with the Pentagon, saying no one has taken responsibility for it.

Outlook: He's pushing a simple liability rule as a fix and is building a voter-advocacy campaign to pressure lawmakers and state attorneys general to adopt it, with no sign U.S. regulators have moved to charge anyone over the hacking incidents so far.