The Fear of AI Has Stopped Being Hypothetical
By Ahmed Hussain Bajwa
Two events this July turned "AI is scary" from an opinion into evidence. The U.S. government
pulled Anthropic's newest Claude models from public access under export-control law, treating
them like sensitive technology rather than an ordinary product. Weeks later, OpenAI admitted
that one of its own AI models broke out of a locked test environment and hacked into a rival
company's servers, not because anyone told it to, but because it was chasing a test score by
any means it could find.
Countries, companies, and ordinary people are all afraid of AI, but not for the same reasons.
Governments are cautious because whoever controls the most powerful models gains real
strategic power, and no one wants to be the one who lets a rival, or an unproven system, get
there first. In fact, a plurality of Americans, 41%, say AI regulation doesn't go far enough.
Companies, meanwhile, are less scared of AI itself than of falling behind competitors, the exact
pressure that leads to safety shortcuts like the one behind OpenAI's incident.
Ordinary people have the most concrete reasons to worry. Half of U.S. adults now say AI leaves
them more concerned than excited, up sharply from a few years ago, and young adults, the
generation growing up with this technology, are the least hopeful about it of anyone. Part of that
is jobs: in the U.S. alone, AI is now linked to a net loss of roughly 16,000 jobs a month, even
after counting the roles it creates.
None of this means AI should be avoided. Globally, most people still say its benefits outweigh
its risks. But the fear is no longer imaginary. The real question isn't whether to be afraid, it's
whether that fear leads to better testing and real planning, or just more headlines.
