AI News Roundup: Meta's $200 Agent, the Bubble Talk, and Apple's Big Week

Five things worth your time today, none of which need a computer science degree to follow. We cut the hype and tell you what each one means for the way you run things.

In a nutshell: Wall Street got the jitters about AI spending, Meta wants to charge you $200 a month for an AI assistant, OpenAI built a safety setting for people who handle sensitive data, Apple is about to show its hand on AI, and Meta got caught running an AI clickbait machine. Here is the plain-English version.

1. The market got nervous about AI

Flat vector illustration of a wobbling bar chart balanced on a single thin column, coral accent, cream background, suggesting a fragile market

The mood around AI investment turned sour. The Nasdaq had its worst day in more than a year, and the chip index dropped over 10 percent in a single session. Part of the trigger was Broadcom, which posted strong AI growth but did not raise its longer-term outlook, and that was enough to spook investors.

Here is the worry underneath it. Companies are pouring huge sums into AI, with US capital spending projected to top 500 billion dollars across 2026 and 2027. But actual consumer spending on AI services sits at around 12 billion a year, and economists now think most industries will not feel real productivity gains for another two years.

What this means for you: do not let stock market drama push you into rushing or panicking. The gap between AI spending and AI payback is real, so pick projects with a clear return you can measure this year, not someday.

2. Meta wants $200 a month for an AI agent

Flat vector illustration of a small robot assistant standing beside a price tag, coral accent on the tag, cream background

Meta is planning its first paid AI product, an agent called Hatch, and reports say the top tier could cost up to 200 dollars a month. Hatch is built to do real work for you, like managing your schedule, sending emails, and even building small software tools from a plain description.

There will be a free version and a premium “Hatch Plus” tier with five to ten times the usage limits. A wider US launch is expected in July. That price puts Meta level with Anthropic and OpenAI, who both charge 200 a month at the top end.

What this means for you: the price of a capable AI assistant is settling around 200 a month per heavy user. Before you buy seats, work out who actually needs that power and who is fine on a cheaper tier.

3. OpenAI added a setting to protect sensitive data

Flat vector illustration of a padlock over a folder of documents with a closed gate, coral accent, cream background

OpenAI rolled out “Lockdown Mode” for ChatGPT. It is an optional setting aimed at people and organisations that handle sensitive information and want stronger protection against a trick called prompt injection, where hidden instructions on a webpage or in a file try to make the AI leak data.

When you switch it on, ChatGPT turns off the riskier features that connect it to the outside world, like live web access, agent mode, file downloads, and connectors. OpenAI is honest that it is not a perfect shield, but it cuts the chance of your data walking out the door.

What this means for you: if your team feeds client records, contracts, or financials into ChatGPT, this setting is worth knowing about. Fewer open doors means less risk, even if it costs you some convenience.

4. Apple shows its AI hand on Monday

Flat vector illustration of a glowing speech bubble emerging from a smartphone, suggesting a voice assistant, coral accent, cream background

Apple’s developer conference starts Monday, and the headline act is a rebuilt Siri. Reports say the new Siri will be powered by Google’s Gemini models, behave more like a proper chatbot, and live in the Dynamic Island so you can swipe down and start a conversation anywhere on the phone.

Do not expect new hardware. This is a software event, with iOS 27 and updates across the Apple range, plus more polish on the “Liquid Glass” look introduced last year. The real story is whether Apple can finally make its assistant useful.

What this means for you: if your business runs on iPhones, a smarter Siri could change how your team handles quick tasks and dictation. Worth watching Monday before you commit to other voice tools.

5. Meta got caught running an AI clickbait feed

Flat vector illustration of a newspaper page dissolving into scattered pixels, coral accent, cream background, suggesting AI-generated content

Meta’s standalone AI app had a “For You” feed that quietly churned out fake news stories, with the headlines, images, and full articles all generated by AI. One example was a made-up piece about a royal butler settling the “milk first” debate. The feed had been live for months.

After The Verge started asking questions, Meta said it would pull the feature. It is a small story with a big lesson about how easily AI can produce convincing rubbish at scale.

What this means for you: treat AI-generated content as a draft, never a finished product. If a platform this size can ship fake stories by accident, your own AI output needs a human check before it reaches a customer.

The bottom line

The theme is the gap between AI promise and AI proof. Money is being spent fast, prices are climbing, and the tools are getting more capable, but the real payoff still needs careful, measured decisions on your part. Spend where you can see a return, protect your data, and keep a human in the loop.