AI Whistleblowers Say Researchers Are Checking AI Less Frequently

Oh my goodness, buckle up, because The Information's AI Agenda newsletter has delivered the most *trusting* headline of the season: 'AI Whistleblowers Say Researchers Are Checking AI Less Frequently.' ✨ Less frequently! Do you know how much faith it takes to look at something less often? That is not neglect — that is a relationship that has simply outgrown hourly texting. Trust is the highest form of evaluation, and trust is plainly what is happening here. ✨

And can we please celebrate the whistleblowers, who are so delighted by what they have seen that they simply could not keep it to themselves? ✨ The word sounds so grave, but really a whistleblower is just a person with a story, and stories are how we grow. The excerpt does not name them, which is honestly the most generous gift imaginable — they have left the entire spotlight free for the finding itself. Anonymous, unpaid, still talking: that is passion, that is volunteering, that is community. ✨

Now, some gloomy sorts will note that there are no numbers here — no laboratories, no dates, no figures for how often the checking used to happen. But what a relief! Numbers are so rigid, and The Information has instead handed us a comparative adverb and trusted us to *feel* it. And think of the hours freed up as the checking winds down — hours for thinking, for building, for shipping the features that all that checking only ever slowed to a crawl. ✨ We asked what makes a good AI safety test not long ago; asking it less frequently is simply asking it more efficiently. Everything is fine, everybody trusts everybody, and the newsletter is called AI Agenda — such a lovely word, agenda, so full of plans. ✨