Corporate policies, safety frameworks and the institutions trying to steer powerful AI systems.
310 articles
Enterprises are finding that narrowly scoped AI agents with clear rules outperform fully autonomous systems in production. As costs, unclear ROI, and weak governance threaten agentic AI projects, competitive advantage is shifting toward trust, auditability, and approval by risk and compliance teams.
VentureBeat · 28d ago
Shanghai Intellectual Property Court upheld a RMB 100,000 fine against a company that deliberately manipulated Baidu’s AI search associations to fabricate evidence for a trademark infringement lawsuit. The ruling is described as China’s first civil judicial sanction for malicious evidence gathering through AI hallucinations.
虎嗅 · 29d ago
This arXiv paper introduces bounded sovereignty, a framework for evaluating AI oversight when organisations use models through managed APIs without owning the weights or infrastructure. Its synthetic experiments show that logs, intervention gateways, trace access, and model-version controls materially affect safety, diagnosis, and post-incident accountability.
ArXiv AI · 29d ago
Odense-based startup Velatir has raised €5 million in seed funding six months after its pre-seed round. The company provides visibility and control over the AI tools employees are already using, with Spintop Ventures and Ugly Duckling Ventures co-leading the round.
The Next Web (TNW) · 30d ago
This position paper argues that AI leaderboards underserve the Global South because of weak governance, conflicts of interest, and inflexible metrics—not a lack of data. Using India as a case study, it recommends independently governed regional leaderboards that incorporate benchmarks such as IndicSUPERB, MILU, and LAHAJA.
ArXiv AI · 30d ago
A position paper analyzing 500 Hugging Face model cards argues that model cards alone do not provide enough information for governing open-weight foundation models. It proposes combining model cards with acceptable use policies and licenses to address safety, provenance, behavior, and enforcement gaps.
ArXiv AI · 30d ago
OpenAI is seeking to compete more directly with Anthropic by introducing stronger privacy protections for enterprise customers. The two companies are increasingly competing over how securely customer data is handled.
TechCrunch AI · 31d ago
OpenAI is launching an initiative to strengthen democratic oversight of AI used in national security. The effort will support government institutions with tools, training, and expertise.
OpenAI News · 32d ago
Gemini can access data across Gmail, Docs, Calendar, Chat, and other Google Workspace services by default. Administrators should review the setting and disable access if it does not fit their organization’s privacy or governance requirements.
ZDNet AI · 32d ago
Analysts warn that vibe coding—building applications through conversations with generative AI—can expose hard-coded secrets, sensitive files, and intellectual property. Its speed and accessibility may also encourage employees to create large numbers of applications without consistent reviews, scanning, or security controls.
Computerworld · 32d ago
This position paper argues that AI governance should complement jurisdiction-specific laws with ISO-like interoperability protocols. It proposes machine-readable AI nutrition labels covering bias, energy usage, and data provenance to simplify cross-border compliance and improve public trust.
ArXiv AI · 32d ago
Money Forward launched Cloud AI Token Management, a business service that centralizes usage and cost management for AI services such as ChatGPT and Claude. The service is free for users who obtain a Money Forward ID.
ITmedia AI+ (日本) · 33d ago
The article describes Rust-related rules for using AI in programming: AI may assist with code inspection but should not replace human authorship. It also signals a usage limit that can trigger a “circuit breaker” when AI assistance becomes excessive.
InfoQ中国 · 36d ago
企業正在大量試用 AI Agents，但局部提效尚未穩定轉化為成本、收入或利潤。實踐者認為，企業必須選擇可量化的高價值場景，並將少數 AI 超級個體的方法沉澱為可複製的組織能力。
虎嗅 · 36d ago
AWS presents a multi-agent M&A due diligence system built on Amazon Bedrock AgentCore. The reference architecture combines agent orchestration, knowledge retrieval, and governance controls, with a complete sample deployable in an AWS account.
AWS Machine Learning Blog · 37d ago
Legal experts say AI agents themselves cannot be held legally responsible for harm they cause. Responsibility generally falls on the person or organization deploying the agent, while developers may also face liability depending on the circumstances.
The Guardian Technology · 38d ago
AWS presents a production reference deployment for Claude apps gateway, a self-hosted governance layer connecting Claude Code and Claude Desktop with Amazon Bedrock or Claude Platform on AWS. The guide covers architecture, enterprise deployment patterns, costs, and implementation resources.
AWS Machine Learning Blog · 39d ago
An xAI co-founder is pursuing an open-source AI startup focused on making models trainable and adaptable by anyone. The effort aims to give developers and organizations more control over how AI is shaped for their needs.
New York Times Technology · 39d ago
Anthropic is intensively meeting prospective investors to build confidence ahead of a potential IPO that could become the largest in history. The company is promising continued rapid growth and new measures to address rising public concerns about AI.
cnBeta (Full RSS) · 39d ago
Flow-by-Flow proposes governing AI output in high-loss domains without requiring humans or AI systems to judge whether each item is correct. It uses formal cognitive-cost scoring, institutional capacity caps, identity-bound friction, and bans on batch clearance to control supervisory load.
ArXiv AI · 39d ago
The South Australian government plans to hold a royal commission into artificial intelligence later this year. The announcement signals a forthcoming formal examination of AI-related issues in the state.
iTNews Australia · 40d ago
Copado’s survey found that nearly 90% of Salesforce customers view operational siloing and dependence on individuals as a challenge. More than 80% expect AI to standardize work, but organizations already using AI report a lack of validation processes as a new obstacle.
ITmedia AI+ (日本) · 40d ago
This systematic review examines whether Explainable AI can operationalize the EU Right to Explanation under GDPR Art. 15(1)(h), AI Act Art.
ArXiv AI · 45d ago
Serverworks launched a service that helps organizations establish rules for generative AI use in AWS environments. Its plans include an option to complete the guidelines in as little as one month.
ITmedia AI+ (日本) · 45d ago
The article argues that AI should continuously surface risks and challenge organizational drift, rather than become the final decision-maker. It recommends separating AI’s power to detect and explain risks from the authority to approve irreversible actions.
虎嗅 · 45d ago
Amazon discovered multiple cases where AI deployment mistakes and weak cost controls caused severe budget overruns. One Claude Sonnet project spent $1.8 million, exceeding its budget by 860%, while other projects also incurred unexpected costs.
IT之家 · 46d ago
China’s National Radio and Television Administration reported removing more than 13,300 non-compliant AI-modified videos and handling over 30 accounts in July. The program targets AI-generated content that distorts classic works, history, cultural symbols, or children’s animations.
IT之家 · 46d ago
The UK is considering proposals that would require employers to ask workers before deploying bossware and other workplace-monitoring tools. The measures could cover AI productivity scoring, keystroke logging, biometrics, and related surveillance technologies.
The Register - AI/ML · 46d ago
National Australia Bank will soon test the security and operational guardrails of an agentic AI platform. The initiative signals that NAB is moving toward broader deployment of AI agents for banking customers.
Bloomberg Technology · 46d ago
Google DeepMind CEO Demis Hassabis warns that we are in a critical window before the arrival of AGI. He emphasizes the need for urgent safety measures to ensure control over advanced AI systems.
TechRadar AI · 57d ago