Two PHP flaws were found with the help of AI last week. Companies, including my own, scrambled to harden their websites. While others, including banks, airlines, and governments had to pull their websites offline. An Anthropic security executive said it will get worse before it gets better. Fixing flaws, reviewing agent output and trusting a teammate over a chatbot are all, still, human work. At the same time, with inflation hovering at 3.7% following the Fed’s first rate hike in three years, companies are being forced to pay for an increasing AI bill by shrinking payroll.
This week’s coverage:
AI Found the Flaws. People Still Have to Fix Them. Some of the bugs sat undetected for over a decade, and then the humans inherited the backlog.
Microsoft Wants You to Trust the Models Least Microsoft says frontier models can’t be trusted and wants to sell you the watchdog.
The Budget Moved First. The Job Redesign Didn’t. Barclays cut 5,000 jobs for AI, 89% of U.S. CEOs said they expect to add headcount, and only 4% of CIOs have plans to redesign roles.
More Meetings, Less Trust, Fewer Juniors Anthropic’s calendars filled up, teams started leaning on bots instead of each other, and nearly half of employees admitted overstating their AI skills.
Here’s what I was reading.