The Informed Layer · Newsscan
Daily briefing

AI execs face lawmakers; OpenAI turns on ChatGPT ads

← All issues

Anthropic / Claude ecosystem

Claude Code 2.1.289 closes second route past rm guard

Claude Code 2.1.289 fixes a second bypass of the deletion-command guard, hardening sandbox safeguards. Stable remains four builds behind the latest release.

Claude Code quietly wipes session logs after 30 days

Claude Code performs a background sweep that deletes session logs after 30 days by default, and the retention period cannot be set to zero. Users needing longer audit trails must export logs manually.

Anthropic adds opt-in voice data sharing for Claude training

Anthropic introduced an opt-in option allowing voice chats with Claude to be used for training, with independent controls for voice versus text data and a data deletion feature. The move aims to improve Claude's speech understanding.

Frontier model providers

OpenAI introduces visual ad format and expanded measurement in ChatGPT

OpenAI announced a new visual advertising format for ChatGPT alongside expanded measurement capabilities and partnerships to gauge ad impact. This marks OpenAI's formal push into AI-surface advertising.

OpenAI details approach to EU text provenance and watermarking rules

OpenAI published its approach to EU text provenance rules, rolling out invisible machine-readable text watermarking (textGrain) for ChatGPT and Codex outputs starting in the EU, with opt-in for API users.

ChatGPT Finances now free for US Free and Go users

ChatGPT's Finances feature is now available at no cost to US Free and Go tier users, expanding access beyond Plus/Pro subscribers. The tool can view and analyse users' financial data.

Meta opens Agent Muse to DIY developers via embeddable integration

Meta is extending Agent Muse to DIY developers through an embeddable integration for third-party apps and hardware, building on the Muse ecosystem.

Cohere launches North 2 enterprise agent platform with memory and budget controls

Cohere unveiled North 2, adding agent memory, granular admin controls, shared libraries, reusable agents and budget-based orchestration for enterprise AI workflows.

Mistral launches multilingual AI model for European public administration

Mistral introduced a multilingual AI model tailored for EU public administration, focused on data privacy compliance and document processing for government use.

GPT-6 Astra sets new ARC-AGI-3 benchmark record

OpenAI's GPT-6 Astra achieved a leading score on the ARC-AGI-3 benchmark under the Standard harness, indicating strong cross-domain problem-solving and efficiency.

AI developer tooling & infrastructure

No significant new developments.

Cloud & platform providers

AWS Capabilities by Region adds feature-level availability notifications

AWS Capabilities by Region now offers per-feature availability notifications with advanced filters, in-app alerts, and weekly digests for region-based tracking.

AWS launches Digital Sovereignty Lens for the Well-Architected Framework

AWS introduced a dedicated Digital Sovereignty Lens for the Well-Architected Framework to evaluate sovereign deployments and guide governance decisions.

Cloudflare Traces enters open beta with automatic end-to-end request visibility

Cloudflare launched Traces in open beta, providing automatic end-to-end tracing across the platform and unifying previously separate logs into a single timeline.

Cloudflare AI Search moves to GA with November 1 billing at $0.75 per million tokens

Cloudflare AI Search transitions from preview to general availability with new pricing, adding image embeddings and OCR-enabled PDF ingestion capabilities.

Cloudflare launches Workers KV Instant with sub-2ms global reads

Workers KV Instant delivers sub-2ms p99 reads with 250ms global write replication across Cloudflare's edge network, enabling near real-time global configuration data.

AI policy, regulation & governance

Lina Khan calls for law mandating AI testing, backs Sanders-Casar bill

Former FTC Chair Lina Khan called for federal legislation mandating AI testing with penalties, giving public support to the Sanders-Casar bill — a shift from voluntary self-regulation toward enforceable federal mandates.

Federal AI inquiry begins in Australia as Firmus seeks data centre rules and OpenAI seeks forgiveness

The federal AI inquiry started hearings in NSW Parliament, with Firmus Technologies pushing for new data centre rules and OpenAI addressing fallout from the rogue agent incidents, focusing on AI training copyright and infrastructure regulation.

AI executives testify before NYC Council on existential AI risks

Executives from Anthropic, OpenAI, Google, and Meta appeared before the New York City Council, the first such city-level appearance, warning that misaligned AI could end in human extinction and urging a federal framework.

Malaysia to table AI Governance Bill in Parliament in early 2027

Malaysia's government confirmed a cross-sector AI Governance Bill will be tabled in Parliament in early 2027, focused on safety, human rights, and welfare.

Industry & market moves

Cohere and PwC announce global secure enterprise AI alliance

Cohere and PwC announced a global alliance to deploy secure enterprise AI, combining Cohere's platform with PwC's transformation expertise, starting in Canada.

Infineon completes acquisition of C2i Semiconductors for AI datacenter power management

Infineon completed its acquisition of C2i Semiconductors to bolster AI data centre power management capabilities and expand R&D in India.

Collibra acquires Trail ML to add agent-powered AI governance automation

Data governance vendor Collibra acquired Trail ML, adding agent-powered AI governance automation to scale enterprise AI controls across data and systems.

ThreatBook acquires CyberStrikeAI as penetration testing moves toward autonomous security

ThreatBook acquired CyberStrikeAI, integrating an autonomous AI penetration testing platform into its offensive security capabilities.

AI product & feature launches

Iterate.ai's Lifeboat runs up to 6x more AI agent sessions per GPU

Iterate.ai announced that its Lifeboat platform increases concurrent AI agent sessions per GPU up to sixfold using confidential computing and optimised inference to address memory pressure on shared hardware.

Binance launches Binance Intelligence AI suite

Binance launched Binance Intelligence, a unified AI suite with an AI Pro tier offering personalised market navigation tools, available free to all users from October 5.

Research with immediate practical relevance

Horizon releases MF-1 unified multimodal model

Horizon introduced Multimodal Flow MF-1, a single unified framework handling language, vision, and other modalities within one shared representation, potentially simplifying training and extension to new modalities.

Researchers track persistent Chinese AI 'agent fleet' operating across the internet

Researchers are tracking a persistent, broadly apparent Chinese AI agent fleet active across the internet, including potential API rule circumvention, in ongoing monitoring after the Hugging Face incident.

Z.ai releases full 753B FP8 weights of GLM-5.3

Z.ai publicly released the complete 753B-parameter FP8 weights of GLM-5.3, enabling full self-hosting and shifting discussions toward hardware and licensing considerations.

UC Irvine's CORNAV robot planner reads CAD floor plans, lifting task success from 13% to 72.2%

Researchers at UC Irvine demonstrated CORNAV, a robot planner that fuses CAD floor plans and look-ahead schedules with LLM-based hazard scoring, raising construction robot task success from 13.0% to 72.2%.