Monday · 2026

AI roundup

20 stories · 19 outlets · 4 min read

AI CSAM reports surge 8,400% in two years

The day's most consequential story is a grim data point: reports of AI-generated child sexual abuse material have exploded 8,400% since 2023, according to a new study. That figure reframes every other piece of news, from privacy warnings about chatbot chats to calls for data center moratoriums, as part of a broader reckoning with AI's real-world harms.

In today's edition: Harm and accountability, Models and infrastructure and Work and tools, plus quick bites.

Luminpdf, Gabeorlanski, Nytimes and 16 more

Harm and accountability

The day's biggest story landed with a number that's hard to ignore, and it sits alongside pieces asking who answers when AI makes the call.

nytimes.com · 14 min

AI Companies Can Read Your Private Chats. Here’s What to Do About It.

The NYT piece walks through how AI chatbot conversations can be read by humans and handed over to law enforcement. A DuckDuckGo survey found many users share intimate details with bots, unaware of the privacy exposure. Timely given the CSAM report, the same data pipeline that enables abuse also enables surveillance.

supplychainbrain.com

0
AI Decision-Making in Supply Chains: Who Answers When the Algorithm Decides?

A short but sharp piece on supply chain AI: an authority matrix tells the system what it may do, but an accountability map tells the organization who answers when it does it. That distinction matters more as autonomous decision-making spreads beyond chatbots into logistics and procurement.

Models and infrastructure

On the technical side, open weights keep closing the gap with closed models, while universities and provinces wrestle with the physical costs of compute.

techtwitter.com

0
Open weights Qwen 4 beats Fable 5 on local hardware

The claim that open-weights Qwen 4 beats Fable 5 on local hardware, if accurate, means near-frontier capability could run on a local machine instead of through a closed API. That shifts the calculus on cost, privacy, and security, and the author's warning about bad actors gaining these capabilities makes the claim's accuracy worth watching.

Work and tools

The practical consequences of AI adoption continue to ripple through jobs, software engineering, and even summer job applications.

gabeorlanski.github.io · 17 min

0
Slop Is Moving Up the Stack | Gabriel Orlanski

The author argues AI agents have mostly solved 'slop' at the implementation level, but the problem has moved up to design decisions, agents rewriting existing libraries instead of using them. The new LibraryDesignBench and LibraryUseBench benchmarks aim to measure this shift.

Quick bites

Sunday

15 Nov's edition

AI · 18 stories