What moved in AI, search and tech. Facts, with the link.

Alphabet loses about $500 billion in market value in one day

Alphabet, Google's parent company, had its worst day ever on the stock market on October 29. Its Class A shares closed down 12.64% at $283.82, erasing about $499.8 billion in market value, according to The Wall Street Journal as cited by Fox Business. The drop followed third-quarter results: revenue rose 25% to $127.6 billion but came in below what Wall Street expected. Google Search revenue grew 12% from a year earlier, after 19% in the first quarter and 17% in the second: search is still growing, but more slowly. Alphabet spent $51.4 billion on data centers and computing in the quarter, and Google Cloud more than doubled to $30.7 billion. Coverage ties the drop to that slowdown in search and to how AI is changing the way people search. Not financial advice.

The Switch 2 is reportedly jailbroken, with help from AI

A jailbreak is an unlock that lets a device run unauthorized software. Security researcher Sammy Hajhamid published what he calls the first jailbreak for the Nintendo Switch 2. It runs through SmileBASIC 4, a game that has since left the eShop, and it is lost every time the console turns off. He says he built it in about two weeks with help from OpenAI Codex, an AI coding tool. The outlets we read have not verified it independently, and Nintendo has not commented.

For information only: we do not encourage using these tools or piracy. We share this story from a research and news perspective.

Critical PHP flaws: time to update

PHP, the language that runs WordPress and many websites, fixed two critical flaws that let attackers run code on a server without logging in. The fix ships in versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and PHP urges everyone to update now; 8.1 and older no longer get official fixes. Security firm Calif found one of the flaws with an OpenAI AI model, and Wiz reported both independently. It is not an isolated case: in April, Anthropic reported thousands of flaws found with AI in operating systems and browsers, now being fixed.

ChatGPT ads arrive in Colombia

OpenAI expanded its ads to 26 more countries, including Colombia, Argentina, Chile and Peru, for 89 in total. They appear only on the Free and Go plans, are labeled and kept separate from answers, and OpenAI says they do not change what ChatGPT answers. More than 100,000 advertisers have used them.

Stolen API keys and AI models without safeguards are sold on the dark web

Security firm Flare analyzed more than 100,000 Telegram channels and 100 dark web forums. Mentions of open AI models with their safeguards removed went from 11 to 93 per 30-day window between January and October 2026. Most of what is sold looks like resold or stolen access, using other people's API keys. Flare documents seven AI-assisted attacks between May and October; one reached at least nine South Korean banks. The study warns that its counts are a floor, not a census. For a business: protect your API keys and review usage on your AI accounts.

The PS5 faces a wave of jailbreaks

A jailbreak is an unlock that lets a device run unauthorized software. Since late September, a public one has circulated for PS5 firmware 7.00 through 13.60. Sony answered with versions 14.00 and 14.10, and this week a developer privately showed access to 14.10 without releasing it. Some outlets say AI was used to analyze the console; the reports we read do not confirm it. Sony has not commented.

For information only: we do not encourage using these tools or piracy. We share this story from a research and news perspective.

Anthropic commits 5% of its shares to a public benefit fund

Subject to stockholder approval, the fund can add 2% more for every 50% increase in the company’s value above a set baseline. Anthropic says it could use it, for example, to address AI’s effects on jobs. Combined with earlier pledges by founders and employees, roughly 20% of its shares will go to the public good.

Citizen Lab confirms online ads that installed spyware without a click

Ads bought by an advertising company infected at least 45 devices at the Al Jazeera network in 2024 and 2025, without anyone clicking. The infection was targeted, but that company's ads appeared on almost 200,000 websites and apps. Google and other platforms cut off its access. The report recommends blocking ads on work devices.

Google adds more links to the web in its AI answers

AI Mode and AI Overviews will show inline links higher up, news carousels in more languages, and related links with images. Google says use of Preferred Sources has grown more than 7x this year.

AI is making it cheaper to rebuild commercial software: the Adobe case

Engineer Brandon Thomas used Anthropic's Claude Opus 5.5 to build ArtCraft, free and open versions of programs like Photoshop, Word and Excel; according to VICE, they are still quite buggy. He says he started from public file-format specifications; an intellectual property expert says it is unclear whether that method holds up to legal review. On October 13, HSBC downgraded Adobe from "buy" to "hold," citing the risk of free AI-built alternatives; its analyst also flagged security risks in AI-generated code. Meanwhile, REA, an open tool that connects AI agents to reverse-engineering programs, passed 300,000 stars on GitHub.