AI News for October 18, 2026

A weekly AI brief rounds up model launches from Meta, NVIDIA, Vivix, and others, alongside infrastructure and workplace-adoption debates from Accenture, Microsoft, and the Patel-Cowen discussion.

Anthropic released Claude Code v2.1.302 with an official plugin repository, a JavaScript/TypeScript Mods system for customizing agent behavior, and expanded cloud sessions that let coding tasks run beyond the terminal.

Perplexity released Decider v1.2, a 27B judgment-probability model plus a 4B Flash variant, adding joint multi-question requests and vision support with benchmark gains that still need third-party validation.

A weekly newsletter roundup covering major AI releases from Anthropic, Google, Meta, OpenAI, xAI, NVIDIA, and others alongside a curated list of new AI research and reports.

A Law360 landing page for its AI legal coverage and newsletter, listing article teasers on legal operations credentials, a Super Micro export-control suit, and a New Jersey deputy AG charged over AI-generated child pornography.

A reference page from Price Per Token listing recent AI model releases and their per-token pricing across providers like OpenAI, Anthropic, Google, Qwen, and others.

The Guardian's AI topic landing page aggregates recent headlines on AI models, datacenters, security, regulation, and workplace issues.

Gallagher Re's 2026 Global InsurTech Report shows funding rebounding to $2.44bn in Q2 2026 with AI-focused startups capturing 99% of investment, while warning that rapid AI data center construction is straining insurers assessing infrastructure risk.

Nvidia's $500 billion financing push with major financial partners raises questions about whether AI chip demand is genuine or inflated by circular financing arrangements.

Indian AI startups raised about $50.6 million of a $302.85 million weekly funding total, led by RapidCanvas's $29 million Series B, as capital keeps flowing to applied AI and agent tools rather than foundation models.

The extracted publisher text is a FinancialContent stocks page aggregating third-party market headlines on AI chip stocks, bank earnings, and tech equities, not the titled article on AI in finance IT services.

Simply Wall St markets three stocks from its Global Hyperscale Cloud and Data-Center Operators screener as beneficiaries of shifting AI infrastructure demand after DeepSeek's latest model release.

A detailed assessment of cyber and AI-driven threats to election and campaign officials ahead of the 2026 US midterms, emphasizing trusted-identity exploitation, OAuth consent phishing, and the gap between institutional responsibility and operational access control.

A Wesleyan Media Project study found 220 AI-generated or AI-enhanced political ads ahead of the 2026 US midterms, backed by over $137 million in spending and dominated by Republican sponsors, with only 27% disclosing AI use.

An educator argues that student AI misuse exposes deeper flaws in assessment systems that reward final answers over genuine understanding, and that AI could improve access to tutoring and feedback.

A weekly SheWritesAI community digest curating member articles on AI, creativity, ethics, and writing, including collaborations and new writer highlights.

The Guardian's Schools section page aggregates recent education articles covering arts access, playground protests, student activism, and special-needs funding.

A daily AI news roundup covering Samsung tapping TSMC 3nm for Nvidia HBM base dies, Anthropic's outsider input on Claude values, Promise Robotics' D.R. Horton deal, and an eighth-day semiconductor selloff.

A general explainer compares NPUs and GPUs, outlining their architectural differences, power efficiency tradeoffs, and the shift toward hybrid on-device AI in consumer hardware.

Uvik Software's vendor blog lays out a framework for measuring AI ROI, arguing that productivity gains only become financial value when they reduce spending, avoid hiring, boost profitable output, or prevent losses, citing evidence of a gap between individual productivity and enterprise-level impact.

A sponsored b2b feature covers Luma's 2026 survey of 760 creative professionals, finding 81% have released AI-assisted content while 62% worry about overreliance on AI.

A detailed comparison of Perplexity, ChatGPT Search, and Google AI Mode across pricing, underlying models, citation accuracy, and real-world adoption as of October 2026.

A Teacher Tapp analysis comparing 2023 teacher predictions with 2026 reported behavior finds AI use for lesson planning and parent communication outstripping forecasts, while marking and attainment-data analysis fall far short amid teacher caution about delegating student judgement.

This is a Guardian Health section index listing numerous health-related headlines spanning NHS policy, trans healthcare, pollution, and public health, rather than a single news article.

This week's security roundup covers mass hacks targeting European supermarket refrigeration units, an Ahold Delhaize cyberattack outage, and an OpenAI insider charged with planting a recording device in an executive conference room.