Anthropic Proposes 'AI Safety Act' to Congress Amid Government Sanctions
Anthropic, currently under U.S. government sanctions, has proposed draft frontier AI safety legislation to federal lawmakers.
Read the original →Article digest · Oct 18 PM
AI Regulation, Safety & Geopolitics — what each story says
Every story in this issue with a short summary, so you can decide what to read before leaving the site. Summaries are written by AI (Claude) only from each publisher's own description and checked by Jev for anything not in that description. They are not full translations. Follow "Read the original" for the full story.
13 stories10 with summaries5 topics
🧪 Safety research and evaluations4
Bill Gates leans on CEOs as he plans a bigger AI safety push, Axios reports
Bill Gates told one person he consulted on AI safety, "I can reach CEOs on this better than you can," Axios reports.
Read the original →FT: Anthropic wants lawyers’ negotiating ethics built into Claude agents that bargain against other agents
Anthropic is reportedly working on giving its autonomous Claude agents built-in negotiating ethics. The development aims to improve the agents' bargaining and dispute resolution capabilities.
Read the original →- Headlines only (the publisher did not provide a description)
- Tencent says third-party models, not Hunyuan, responsible for AI agent activityChina Daily · 10/18 05:22 UTC
🚨 Incidents and misuse4
Forter Just Shipped the Most Complete Merchant-Side Stack for AI Agent Traffic. Holiday Peak Is Five Weeks Away.
Agentic orders across Forter's network have quadrupled since August. Fraud doubled in two months. The description says the merchant liability wall is not waiting for merchants to get ready.
Read the original →The AI Agent That Reads Your Data May Belong To Someone Else – OpEd
An Indian cyber warning and a worldwide intrusion campaign expose a new accountability gap. Companies can suffer AI-enabled access without deploying AI.
Read the original →AI in production faces silent failures from drift, review bottlenecks
A recent InfoQ webinar highlighted operational challenges in deploying agentic AI systems. Issues like semantic and embedding drift can lead to subtle, hidden errors. Experts discussed new engineering approaches for evaluation, security, and governance.
Read the original →- Headlines only (the publisher did not provide a description)
- DraftKings class action claims AI model targets losing bettorsTop Class Actions · 10/18 05:29 UTC
🔌 Chips and compute2
Quad 8-pin mod protects the $8,500 RTX 5090 you can’t replace from melting — modder’s custom power solution claims to fix Nvidia 16-pin GPU meltdowns
The description states that RTX 5090 production halts make protecting the GPU from meltdowns critical.
Read the original →ASML and TSMC Signal Chip Expansion, but AI Returns Remain Unproven
ASML's equipment sales and installed-base management supported its latest quarter. TSMC kept its 2026 capital spending budget at $60 billion to $64 billion.
Read the original →🤝 International frameworks2
NIST’s 2026 AI-Agent Warning: SMBs Are Missing The Point
NIST flagged AI-agent identity risk for Cybersecurity Awareness Month 2026. Most small businesses are still solving last year's problem, according to the description.
Read the original →- Headlines only (the publisher did not provide a description)
- The AI Trust Crisis in Legal & Healthcare: Moving from “Trusting AI” to “Verifying AI”BBN Times · 10/17 21:29 UTC
🧑⚖️ Litigation and IP1
Family of FSU shooting victim files lawsuit alleging ChatGPT helped shooter plan attack
The family of FSU shooting victim Tiru Chabba alleges in the lawsuit that OpenAI, the maker of ChatGPT, could have done something to prevent the shooting.
Read the original →The analysis and visual summary are in the Oct 18 PM issue. Japanese version here.