This Week in AI Security Scanners, Apple Intelligence and Bubble Fears

Safety Tools Go Mainstream

Four of the biggest names in AI — Anthropic, OpenAI, Cloudflare and Google — have all released AI-powered scanners that search computer code for security flaws. What makes this notable is not just that they exist, but that they all use the same basic design idea: separating the job of finding bugs from the job of reporting them. This shared approach aims to cut false alarms, which waste developers' time chasing ghosts instead of fixing real problems.

The scanners produce reports in SARIF, a standard file format that lets security tools share their findings with platforms like GitHub. A developer using Claude Code could scan an entire repository and receive a report listing each vulnerability's severity, category, impact and steps to reproduce it. Cloudflare released its security audit skill under a MIT licence, meaning any team can use it with any capable coding agent. The result goes through six phases before reaching a final report.

This matters because every company that writes software needs to keep it safe from hackers, and unreliable scanners are a frustrating roadblock.

Apple Brings AI to Your Wrist

Apple has added two new AI-powered voice features to the Apple Watch in its watchOS 27.3 beta release. The features are called Siri Recap and Live Rewind, and they let the watch listen to conversations, summarize them and replay the last 15 seconds of audio. They work only on the new Apple Watch Series 12 and Ultra 4, paired with a compatible iPhone, and run on Apple Intelligence, Apple's own AI system.

If you've ever missed part of an announcement at an airport or nodded along in a meeting without catching everything, this could be useful — as long as you own the right hardware. The features were announced back in September alongside the new watches and are now available for testing.

AI Hits the Factory Floor and the Living Room

Smarter AI is leaking out of chat apps and into everyday objects. Arrowmax launched a Kickstarter for the Arrowspace S1, a smart electric screwdriver with a built-in camera and an on-device AI model that identifies screws and sets the correct drilling power automatically, so users stop stripping tiny fasteners. Netflix is testing a feature called AI Discover that guesses your mood and suggests movies or shows so you spend less time scrolling and more time watching.

Meanwhile, IBM laid out a new operating model for businesses wanting to run AI agents that act autonomously across a company's data and systems. Instead of waiting for prompts, these agents continuously monitor, decide and execute tasks. IBM says enterprises need four coordinated systems — covering governance, infrastructure, orchestration and evaluation — to make this work reliably.

Money Questions and Model Hacks

On the business side, Alphabet shares fell sharply last week as investors worried about a potential AI bubble. The concern centres on circular financing: big AI companies invest heavily in startups, those startups then spend that same money buying cloud services from the big companies, creating a loop that some fear is inflating values without real economic substance. The Wall Street Journal has flagged the risk to the wider economy if that cycle breaks.

In research, MIT scientists used OpenAI's GPT-6 Astra to rewrite the internal numbers of a smaller model, Qwen3.5-0.8B, boosting its math skills without any traditional training. And separately, researchers auditing early-stopping methods found that letting large reasoning models continue past the point where answers seem ready can actually make them less reliable on math, science, coding and factuality tasks.

Finally, Anthropic updated Claude Code to refuse prompts asking it to recreate someone else's video games, closing a loophole that had been used to copy and port console titles to PC.

The week in short

• Four AI labs — Anthropic, OpenAI, Cloudflare and Google — shipped security scanners using the same shared design
• Alphabet lost over half a trillion dollars in market value over AI bubble concerns
• Apple Watch Series 12 and Ultra 4 get Siri Recap and Live Rewind in watchOS 27.3 beta
• MIT researchers made Qwen3.5-0.8B far better at math using GPT-6 Astra with no training
• Itch.io saw daily game listings jump from 3,000 to 8,000 in two months as AI tools flooded the platform

Every story this week

• Warm words make ChatGPT , Grok , Gemini act like hurt AI, study finds
• Anthropic, OpenAI, Cloudflare and Google ship AI security scanners that work the same way
• OpenAI merges ChatGPT's chat and Work tabs into one screen
• Alphabet stock drops as investors fear the AI bubble could pop and hurt the whole economy
• Arrowmax Arrowspace S1 uses AI and a camera to find the right screw and set the right power
• Nace AI opens Drex Max v1, a 27B decision model that runs on your own Mac
• Researchers propose Avatar Intelligence to make AI agents better at handling your daily tasks
• Itch.io sees AI game pages double as generative tools flood indie store
• Apple Watch gets AI features to summarize conversations and replay what you missed
• GPT-6 Astra fixes Qwen3.5’s math ability by rewriting its brain without any training
• IBM puts AI agents at the center of a new operating model for enterprises
• Researchers audit early stopping in AI reasoning models, find answers can get worse further along
• Anthropic blocks Claude Code from rebuilding existing video games
• Netflix uses AI to help you stop scrolling and start watching