Meta $25B Penalty, PHP Flaws, Legora: Top Stories Today

Six big stories made waves in tech over the last 24 hours. A judge ordered Meta to pay $25 billion over data privacy. AI tools found two critical flaws in PHP, the code behind much of the web. And legal AI startup Legora raised $550 million at a huge valuation. Here is everything you need to know, in plain words.

Meta Ordered to Pay $25 Billion in Privacy Case

A judge in New Mexico ordered Meta to pay $25 billion on Wednesday. This came after a jury found the company misled users about data privacy. It is one of the largest financial penalties against a company in US history.

The case goes back to the Cambridge Analytica scandal. The British firm, which worked on Donald Trump’s 2016 campaign, got personal data from up to 87 million Facebook users through a third-party app. New Mexico sued Meta in 2021, saying it lied to users about who could access their data.

The jury found Meta broke the state’s consumer protection law more than 43 million times. The judge said over $219 billion was justified under state law but cut it to $25 billion. Attorney General Raul Torrez said the fine, which funds public education, could pay for “the construction of schools for the next 100 years.”

Meta will appeal. Chief Legal Officer C.J. Mahoney said the ruling makes it “all but impossible to do business” in New Mexico. In 2019, Meta paid the FTC $5 billion over the same scandal.

Why it matters: This ruling shows state governments can hit tech giants with massive fines. Other states may now bring similar cases, which could force big platforms to be more honest about privacy and data use.

AI Found Two Critical PHP Flaws, Patches Are Out

The PHP project shipped emergency security fixes on October 21 for two serious flaws, tracked as CVE-2026-109618 and CVE-2026-109619. These are memory safety bugs rated up to 9.2 out of 10 in severity, and they can let attackers run code on a server remotely.

Security firm Wiz said it discovered the flaws using its Wiz Atlas AI agent along with Gemini 4 Argon and GPT-5.6 Sol. The bugs affect releases going back to PHP 5.4, which means a huge number of websites could be at risk. OpenAI added that its partner Calif, part of the Patch the Planet program, found one of the flaws with GPT-5.6 Sol. The US National Security Agency even advised taking servers that cannot be patched offline.

The trouble started when details of the flaws leaked before the planned disclosure date. Security firms reported attempts to exploit the bugs, some of them before the fixes were ready. PHP runs a large share of the world’s websites, so this is a race between attackers and site owners.

Why it matters: This is the clearest example yet of AI finding dangerous bugs faster than human defenders can fix them. If you run a PHP server, update it now. It also shows AI security research is becoming a double-edged sword.

Legora Raises $550 Million at $8.8 Billion Valuation

Swedish legal AI startup Legora raised $550 million on Wednesday at an $8.8 billion valuation. That is a 58 percent jump from its $5.55 billion valuation seven months ago. The round was led by Atreides Management, with new backers including Baron Capital, Fidelity, and J.P. Morgan.

Legora calls itself the “agentic operating system for legal professionals.” Its software helps lawyers research, review, and draft documents. Annual revenue is now over $200 million, after climbing from $1 million to $100 million in less than 18 months, then adding the next $100 million in under six months. Bessemer Venture Partners calls it the fastest run to $100 million in enterprise software history.

The platform serves more than 140,000 legal professionals across 2,400 law firms in over 80 markets, and the team has passed 1,000 employees. The United States is its biggest market. Rival Harvey raised $550 million at a $15.5 billion valuation in September, so the legal AI race is heating up.

Why it matters: Legal AI is one of the hottest corners of the AI boom. Law firms are adopting these tools at record speed, and the revenue growth shows they are paying real money for them.

Walmart Opens Its Site to AI Shopping Agents

Walmart announced on October 21 that AI agents will be able to shop on its website starting next month. It built a new system called the Agent Gateway, which connects assistants through the Model Context Protocol, an open standard for linking agents to services. It launches as a beta for developers first, and Sam’s Club gets the same setup.

Walmart also published a rulebook. Every purchase needs your approval by default, but you can set a spending cap, such as $50 or $100. Agents never see your card details and cannot change payment methods. They can only do what you asked. Agent purchases appear separately, you can switch an agent off anytime, and returns apply even when the agent gets it wrong.

The move contrasts with Amazon, which has reportedly blocked agents from Perplexity and Meta’s Muse. Walmart already works with its own assistant Sparky and with integrations from Meta, ChatGPT, and Gemini.

Why it matters: Walmart’s playbook, approval by default, spending caps, no card access, and an off switch, could become the standard for how every retailer handles AI shoppers. Trust may decide who wins agent-driven shopping.

OpenAI Models Arrive on Google Cloud

OpenAI and Google Cloud announced a partnership on October 21 that brings OpenAI’s newest models to Google Cloud’s Model Garden. Three models, GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna, are rolling out now in private preview for select enterprise customers.

Each model serves a different job. Astra handles demanding work like coding, computer use, and scientific research. Sol offers near-Astra intelligence at a lower cost for agentic coding and professional tasks. Luna is fast and cheap, built for everyday work at scale. Developers can also connect local Codex CLI and Desktop workflows to Google’s Gemini platform for running queries.

Businesses buy the models through Google Cloud Marketplace, so everything lands on a single Google Cloud bill and draws down existing spending commitments. Google says it is now the only major cloud provider offering all three leading closed models alongside more than 200 other models. It is the third cloud to sell OpenAI’s models since the exclusive Azure deal ended in April, after Amazon Bedrock and Oracle Cloud. OpenAI stressed that Microsoft remains its primary cloud partner.

Why it matters: The AI model market is going multi-cloud, which gives buyers more choice and simpler billing. Competition like this usually pushes prices down and quality up, which is good for every business using AI.

Anthropic Posts $80.7 Billion First-Half Loss Ahead of IPO

Anthropic’s net loss widened to $80.7 billion in the first half of 2026, according to a draft IPO prospectus reviewed by Reuters. The company is expected to go public in November. The headline number needs context. About $72 billion of the loss was a non-cash accounting charge tied to the rising value of bonds that can convert into company shares, not money spent running the business.

The underlying business is growing fast. Revenue reached $16.3 billion for the six months ended June 30, up from $1.1 billion a year earlier. In the third quarter, revenue jumped 45 percent to $16.8 billion, and the company told investors it was profitable on an adjusted basis.

The costs are still staggering. Anthropic spent nearly $9 billion on computing capacity and infrastructure in the first half, up from $2.3 billion a year earlier. It has committed to spending at least $518 billion over the next decade on cloud services and computing. Some investors are now weighing a lower IPO valuation of around $1.5 trillion, and two customers account for roughly one-fifth of sales.

Why it matters: The accounting charge makes the loss look scarier than it is, but the spending shows how expensive the AI race has become. The November IPO could shape the whole AI market, with OpenAI’s listing possibly following next year.

Sources

• Reuters, CNN, and USA Today on Meta’s $25 billion penalty.
• AI News Online on the AI-discovered PHP flaws.
• Reuters and TechCrunch on Legora’s $550 million raise.
• Gadgets Now, GlobAI, and The Shelby Report on Walmart’s AI shopping agents.
• Google Cloud Blog and Forkast on OpenAI models in Google Cloud’s Model Garden.
• The Straits Times and Channel News Asia on Anthropic’s $80.7 billion first-half loss.