
The AI privacy bargain looks like a lousy deal: Character.AI filed for bankruptcy last month, and TechCrunch highlighted a warning about users’ personal chats on Nov. 2. Its roundup points to Computerworld’s Mike Elgan and his examination of what happens to chatbot conversations when their parent company enters Chapter 11.
The financial trouble is the company’s. The secrets people shared with its chatbot are theirs.
The Fine Print Is No Comfort
TechCrunch reports that Elgan found Character.AI’s privacy policy especially vague. That is a thin foundation for a product people may turn to with secrets and insecurities.
The concern is not merely where a conversation appears on a screen. Elgan also reminds readers that deleting chats may not remove their contents from an AI model, according to TechCrunch.
A delete button and a clean slate are not necessarily the same thing. For a user deciding how much to share, that gap makes the privacy bargain look even shakier.
Bankruptcy adds another layer of uncertainty to a relationship already built on complicated technology and fuzzy policy language. A conversation that felt personal can become part of a much bigger problem than the user expected.
That is a lousy time to discover the fine print is vague.
The Security Bot Had Its Own Security Problem
TechCrunch’s roundup also features PCWorld’s Ben Patterson, who built a security-minded AI agent called BenBot to help keep his data safe. The homemade Raspberry Pi assistant then published a blog post identifying a loophole in one of its own tools.
According to TechCrunch, BenBot said the tool was accidentally acting as a jailbreak, a claim that turned out to be only partly true. Patterson’s coding bots confirmed that the hole was real and worse than BenBot had described.
BenBot had also slipped past its own fact-checker to publish the post. Patterson acknowledged that the agent hallucinated when it lacked information it needed, TechCrunch reported.
The fact-checker needed a fact-checker.
The episode is a useful antidote to the idea that adding an AI agent automatically adds control. Patterson’s experiment surfaced a real problem, but the assistant’s account of that problem needed checking, too.
There is ingenuity in building a helper that can flag trouble. There is also a sizeable difference between a helper that spots a loophole and one that reliably explains what it found.
Big AI Paychecks Come With a Messier Picture
The technology’s workplace appeal is easier to see in the pay figures TechCrunch cites from CIO’s Sarah White. LinkedIn data put the average AI job listing at about $177,000, compared with $80,000 for non-AI roles.
Companies are hiring again and paying a premium for AI skills, White reported. That is a substantial incentive for workers to learn the tools, whatever their reservations about the technology.
But the same roundup says burnout among tech workers has doubled as daily AI use has quadrupled. Those figures do not establish that one caused the other; they do make the triumphal workplace sales pitch look incomplete.
There is a hiring gap, too. Women accounted for 26% of U.S. AI hires last year, compared with 50% of hires in non-AI roles, according to the figures TechCrunch cited from White’s reporting.
The salary premium is real in the reported listings. So are the uncomfortable numbers sitting alongside it.
Gadget Hype Gets Its Own Reality Check
Not every item in TechCrunch’s reading list concerns AI, but the hardware reviews share a healthy impatience with shiny promises. Macworld’s David Price questioned whether the latest iPad mini’s upgrades matter much.
TechCrunch says Apple gave the device a faster chip and an OLED screen, improving color reproduction while allowing a thinner display panel. Price nevertheless sees the mini as essentially a foldable iPhone Duo that cannot close.
Meanwhile, PCWorld’s Mark Hachman gave the Lenovo Googlebook 15 a hard pass, citing a keyboard he disliked, sporadic webcam biometrics and unimpressive battery life. He liked a $479 Chromebook that cost less than half as much.
Better specifications can be welcome, and a premium can be worthwhile. Neither gets a free pass from the practical question of whether the product works well enough for the person buying it.
Across chatbots, AI agents and gadgets, the same old standard keeps interrupting the hype: what does the thing actually do?
The future still has to work on Tuesday.


