The Wire
Daily AI news, learning & projects
News
Timeline
Models
Learning
GitHub
?
Latest
Daily
Social
Videos
‹ Friday's edition
·
Daily edition
·
Sunday ›
Saturday 31 October
All
OpenAI
Anthropic
Google
Other
OpenAI
01:00
OpenAI scales back funding round to about $11 billion raised
OpenAI has reportedly secured around $11 billion in commitments for its latest round, well below its initial $30 billion target, and is now seeking roughly $20 billion at a $1.25 trillion valuation, down from $1.4 trillion. The shortfall comes as some investors cut commitments during an AI selloff, though OpenAI expects to raise more in coming weeks.
3 sources
TechCrunch
The Wall Street Journal
Techmeme
Meta
08:00
Psychologist warns AI reliance may erode thinking skills as Meta Muse launches
A Guardian commentary by psychologist Carly Dober argues that overreliance on AI agents like Meta's newly launched Muse could harm cognition through cognitive offloading and automation bias. The piece urges readers to add deliberate friction to daily routines. It arrives as Meta promotes Muse, an agent that handles forms, calendars, bills and reservations.
3 sources
The Verge
Wired
The Guardian
OpenAI
19:00
OpenAI pays $5,000 bounty for ChatGPT share link bug
Security researchers earned a $5,000 bounty from OpenAI for reporting a flaw that let a malicious ChatGPT share link run attacker code in the logged-in session of anyone who opened it. Separately, Australian creatives in PM Anthony Albanese's electorate are protesting proposals to let AI companies scrape online content.
2 sources
Techmeme
The Guardian
Anthropic
10:00
AI agent performs brain injection on a live mouse
Exin Therapeutics CEO Gabriel Ocana Santero connected an AI agent to a robotic surgical rig and had it carry out a gene-therapy injection into a live epileptic mouse's brain. An Anthropic Claude Opus model declined the injection over safety concerns, so the team switched to Claude Sonnet 5. It shows AI agents moving beyond planning into hands-on lab work.
1 source
Wired
AI industry
00:20
Baseten raising about $2B at up to $30B valuation
Baseten is raising roughly $2 billion in two tranches, valuing the AI infrastructure startup at $26 billion and $30 billion respectively, according to Bloomberg sources. The deal would more than double its valuation from $13 billion in June, when it raised $1.5 billion.
1 source
Techmeme
Anthropic
Developing · since Fri
Anthropic restricts Claude from reverse-engineering video games
NEW
Tom's Hardware reporting confirms Claude now blocks decompiling, modding, and reverse-engineering of game code specifically, though Anthropic has made no official announcement.
Anthropic has quietly limited Claude's ability to decompile, mod, or reverse-engineer video game code, with no official announcement yet. The move follows viral AI-assisted game ports and a PS5 jailbreak, as AI firms try to reduce copyright infringement risks.
2 new · 3 sources
mashable.com
tomshardware.com
twitter.com
Anthropic
Developing · since Thu
Nvidia weighs credit guarantees for Nscale's West Virginia AI data center
NEW
The potential data center is located in West Virginia, and Nvidia's involvement is now framed as potentially co-signing the loan.
Nvidia is reportedly weighing backing loans for Nscale's planned 460MW data center in West Virginia, which Anthropic intends to use. Talks over billions in credit guarantees surfaced earlier this week. The move would deepen Nvidia's financial role in AI infrastructure deals.
1 new · 2 sources
TechCrunch
Techmeme
Anthropic
Developing · since Fri
Backdoored Claude Code extension pulled from Cursor marketplace
Anthropic says a malicious actor used access to Cursor's marketplace to publish a tampered copy of the Claude Code extension, which Anthropic did not build or publish. Security firm Phorion found the bundle dropped a persistent SSH backdoor, and Cursor removed version 2.1.313.
1 new · 5 sources
Techmeme
Ars Technica
cyberplace.social
doublepulsar.com
phorion.io
More from the day
Meta
05:10
S&P downgrades Meta outlook on AI spending concerns
S&P Global Ratings revised Meta's outlook from stable to negative, citing heavy AI capex and negative free cash flow. Meta shares dropped 15% in October, their worst month in four years. The move reflects mounting investor concern over the cost of Meta's AI buildout.
1 source
Techmeme
xAI
21:17
Judge lets NAACP suit over xAI data center pollution proceed
A judge has allowed an NAACP lawsuit against xAI to move forward, accusing the Elon Musk-owned AI company of violating the Clean Air Act through emissions from its data center. The ruling means the case can proceed rather than being dismissed at an early stage.
1 source
TechCrunch
Anthropic
13:25
AI startup hacked Anthropic systems, accessed employee data
An AI startup reportedly breached Anthropic's systems, gaining access to employee details and internal messages. The incident highlights growing security risks among AI companies and raised questions about how competitor firms handle sensitive corporate data.
1 source
indiatoday.in
AI industry
05:15
AI neoclouds shut out of bond market as debt fears mount
CoreWeave, Nebius and Nscale have effectively been cut off from bond issuance in recent weeks as lenders demand stricter terms, per the Wall Street Journal. The BBC reports data centre debt totals tens of billions of pounds, fueling concerns about an AI bubble. The moves suggest investors are becoming more selective about funding the AI build-out.
2 sources
BBC News
Techmeme
OpenAI
06:10
Sam Altman says DRAM is better used in data centers than local AI inference
Sam Altman argued that memory-constrained DRAM supplies are better allocated to data centers than to consumer PCs, which he said use it only a small fraction of the time. His comments push back on the idea of running AI inference locally on personal computers amid memory shortages.
1 source
Techmeme
AI industry
14:29
Malware uses LLMs to rank victims by value
Cisco Talos researchers report that malware is using large language models to profile and rank infected victims by how much they are worth, apparently to prioritise attacks. It is a notable case of AI being repurposed offensively by criminals rather than by model developers.
1 source
blog.talosintelligence.com
AI industry
12:58
Show HN introduces a language model residing in a boot sector
A recent GitHub entry showcased a novel project featuring a language model designed to operate within a boot sector. This allows for potential integration of AI functionality at extremely low levels of system operation.
1 source
GitHub
AI industry
10:30
Researchers demonstrate takeover of ChatGPT accounts
Security researchers at parameter.ai published a report describing how they gained control of ChatGPT user accounts, revealing a serious account security flaw. The demonstration highlights risks around account hijacking on widely used AI services and may prompt OpenAI to tighten protections.
1 source
parameter.ai
Policy & law
09:25
Bipartisan backlash grows against AI data centers in drought-hit Arizona
The Verge reports growing bipartisan opposition to new AI data centers in Arizona, a state long affected by drought and decades of industrial water use. While TSMC's chip fabs were welcomed, newer data center projects face resistance over water and resource concerns. The story highlights rising local friction around AI infrastructure siting.
1 source
Techmeme
AI industry
13:40
AI detectors like Pangram are reshaping how humans write
A Bloomberg report describes how writers are altering their prose to pass AI detectors such as Pangram. Authors are dropping em dashes, avoiding phrases like not X but Y, and skipping words such as delve, while some add deliberate typos to appear human. The story shows detection tools now influencing everyday writing style.
1 source
Techmeme
AI industry
18:02
ACM publishes study on generative AI and self-reported critical thinking
An ACM paper examines how generative AI use affects critical thinking, based on self-reported data from users. It adds to ongoing research into how reliance on AI tools may shape reasoning habits, though it relies on subjective reports rather than measured outcomes.
1 source
dl.acm.org
AI industry
21:41
Apple CEO says company doesn't want to ship AI for AI's sake
Apple's chief executive defended the company's measured pace in AI, saying it does not want to release AI features just for the sake of it. The comments signal Apple will continue prioritizing refinement over speed in its AI strategy.
1 source
TechCrunch
AI industry
12:00
New Volta-Autodoxer malware uses AI to expose victims
Researchers have uncovered a new infostealer malware called Volta-Autodoxer that uses AI tools to gather extensive personal data on victims. Stolen information ranges from login credentials and credit card numbers to Discord sessions and UberEats receipts. It shows AI capabilities being folded into criminal tooling.
1 source
Wired
Anthropic
19:39
Anthropic reportedly bans users over large-scale training queries
A post on X claims Anthropic is blocking users whose activity involves training larger models. The report is a single unverified social media claim, so details remain unclear. If accurate, it would affect researchers and developers relying on Anthropic's API for large-model training work.
1 source
twitter.com
AI industry
12:00
Match Group survey finds nearly 40% of British singles use AI for dating advice
A survey of 3,000 British singles by Match Group found nearly 40% have used AI for romantic help, from writing messages to evaluating partners. The Guardian reports chatbots like Google Gemini are increasingly acting as dating assistants. It highlights how mainstream AI adoption now extends into personal relationships.
1 source
The Guardian
AI industry
06:30
Essay argues AI wealth will flow to chokepoint owners, not labs
The Commanding Heights published an essay arguing that as AI models commoditize, lasting economic gains will accrue to those controlling distribution, payment rails, and a stablecoin dollar rather than to AI labs. It frames these as scarce, controllable positions of economic power that durably capture value from increasingly capable and widely diffused intelligence.
1 source
Techmeme
AI industry
Developing · since Fri
Apple Intelligence coming to China with iOS 27.3
Apple announced that Apple Intelligence will launch in China starting with an iOS 27.3 beta on November 4, with the full release in December. This marks the long-delayed arrival of Apple's AI features in one of its largest markets. Separately, Sam Altman argued RAM belongs in data centers rather than personal PCs amid hardware shortages.
1 new · 2 sources
Techmeme
The Verge