2 new articles showing why you shouldn't trust AI or AI companies. MS Copilot and Perplexity, in these cases.

[highplainsdem]

First, about Copilot, from Bleeping Computer:

https://www.bleepingcomputer.com/news/microsoft/microsoft-says-bug-causes-copilot-to-summarize-confidential-emails/

Microsoft says a Microsoft 365 Copilot bug has been causing the AI assistant to summarize confidential emails since late January, bypassing data loss prevention (DLP) policies that organizations rely on to protect sensitive information.

According to a service alert seen by BleepingComputer, this bug (tracked under CW1226324 and first detected on January 21) affects the Copilot "work tab" chat feature, which incorrectly reads and summarizes emails stored in users' Sent Items and Drafts folders, including messages that carry confidentiality labels explicitly designed to restrict access by automated tools.

Copilot Chat (short for Microsoft 365 Copilot Chat) is the company's AI-powered, content-aware chat that lets users interact with AI agents. ​Microsoft began rolling out Copilot Chat to Word, Excel, PowerPoint, Outlook, and OneNote for paying Microsoft 365 business customers in September 2025.

"Users' email messages with a confidential label applied are being incorrectly processed by Microsoft 365 Copilot chat," Microsoft said when it confirmed this issue.

-snip-

Microsoft wants customers to believe that only the user whose confidential content was read by the AI would have learned of it through the AI agent later, but the comments on the article showed some skepticism:

Bug they say.. yer suuure.
Love seeing stuff like this, because its just another nail in the coffin that is the b$ that is "AI"

"Oops. Sorry. We were just reading those emails for our internal use. We did not mean to show you that we were reading them. It was "an unspecified code error".
Sure it was.



Second, about Perplexity, from MakeUseOf.com :

https://www.makeuseof.com/bought-annual-perplexity-subscription-lied/

If you bought an annual Perplexity subscription, you were lied to
By
Yadullah Abidi
Feb 21, 2026, 7:00 AM EST

-snip-

In practice, Pro users, including myself, enjoyed upwards of 500 to 600 Deep Research queries per day, 50 Labs queries per month, and unlimited file uploads. That is the product people evaluated and committed to $200 for a full year.

-snip-

In early February 2026, Perplexity rolled out an upgraded Deep Research feature powered by Anthropic's Opus 4.5 Thinking model for Max users and Claude 4.5 Thinking for Pro users. The company framed it as a quality improvement, promising "deeper insights and higher-quality results," according to the official changelog. What the changelog conveniently buried was that "usage limits have been adjusted to allocate more computing power per session," which is corporate speak for slashing your quota by over 99 percent.

Here is what Pro subscribers woke up to find. Deep Research went from roughly 500 to 600 queries per day down to just 20 per month, or in some cases (including for myself), only four per month, as claimed by Perplexity's mobile app and web version. Lab queries were cut from 50 to 25 per month.

File uploads, previously unlimited, are now capped at 50 per week. Pro Search limits shifted from clear daily allowances to vague weekly limits that Perplexity's own support staff admitted they can't disclose exact numbers for. If you want anything close to your old usage, you need to upgrade to Perplexity Max at $200 per month—10 times the Pro monthly rate.

-snip-

As I've mentioned in other posts here where I pointed out AI companies are planning to REALLY increase what they charge users, both Sam Altman of OpenAI (ChatGPT) and Perplexity CEO Aravind Srinivas were dreaming out loud last year about being able to charge users $1,000/month in the future. Sam had admitted they were still losing money with their $200/mo subscription tier, because those damn users expected to use that subscription a lot.

Anyone who thinks the AI companies - companies that are losing money with every single use of their badly flawed and hallucinating, but very costly to operate, generative AI tools - will keep offering freebies and cheap tiers is deluding themselves.

As soon as they think you're hooked, you will be price-gouged.

And you will never have control/ownership of the better AI models, because with those requiring so much training on data sets of all the intellectual property the AI bros could steal, and top AI experts commanding 7- and 8-figure yearly salaries, only billionaires can set the rules and prices.

But the AI bros expect you to become increasingly dependent on their products, until a large chunk of your income will have to be spent on AI, with it a priority like housing, food and healthcare.

[SheltieLover]

1. Ty for sharing

[highplainsdem]

2. You're welcome, SheltieLover!

[highplainsdem]

3. kick

[lindysalsagal]

4. I never did trust it. I wish I could rip it out of google.

I tried to turn it all off, but it shows up in searches.

[Intractable]

5. In a limited way, you can rip it out of your searches.

Just add "-ai" to your search string. The AI summary will not appear.

Your search will still feed the AI models, but it's behind the scenes.

[2naSalit]

6. K&R