AI isn’t the threat you think it is: Our laziness is

If you have been following the artificial intelligence (AI) news over the past few weeks, you could be forgiven for thinking we are rapidly approaching the opening scene of a science fiction movie.

OpenAI (ChatGPT) recently disclosed several incidents involving what it describes as model “misalignment” cases in which AI systems behaved in unexpected ways or outside their intended roles. Other AI companies have been confronting similar questions about what happens as these systems become more capable and, importantly, more autonomous. Some of the biggest names in artificial intelligence, like Anthropic (Claude), are now openly discussing stronger safeguards, independent evaluations and how far we should allow increasingly powerful AI systems to operate without human intervention.

These are serious developments and we should treat them seriously. But I worry we are drawing the wrong conclusion from them. AI is not the threat we think it is. Our laziness is. I do not mean laziness in the traditional sense. I mean our tendency as human beings to create powerful tools, rush to use them, and then wait until something goes wrong before establishing the rules. It is our willingness to automate a process without deciding who remains accountable for the outcome. We have this fascination with what technology can do without putting nearly enough effort into deciding what it should be allowed to do.