AI News for July 22, 2026

Google released three Gemini Flash AI models on July 21, 2026, including a cybersecurity variant that autonomously attacks to verify vulnerabilities, alongside cost-cutting and performance improvements.

Poolside announces Laguna S 2.1, a 118B parameter MoE model outperforming larger models on terminal tasks.

This article compares over 50 AI agent tools across different domains, providing practical insights from real-world testing.

Nvidia and AMD are escalating their AI competition by shifting from selling individual chips to supplying complete data center infrastructures, a move that could reshape the market and raise new antitrust concerns.

At the Bridges conference, education leaders showcased real-world AI agent deployments in schools, highlighting differences in autonomy and data access approaches.

The US Army burned through 20 billion AI tokens per day during a 38-day operation, exhausting its annual enterprise pack subscription, despite claims of unlimited tokens.

Google released three new Gemini AI models on July 22, 2026, prioritizing speed and cost-efficiency, while the flagship Gemini 3.5 Pro remains delayed.

Alibaba's stock rises on Qwen 3.8 model preview and Apple partnership news, but the article is a vendor blog promoting TIKR's tools.

Alibaba's Qwen3.8-Max, a 2.4 trillion-parameter MoE model, is previewed with no benchmarks or active parameter count, sparking market interest amid China's AI race.

xAI releases a Grok add-in for Excel that offers AI-powered analysis and financial modeling, requiring a paid subscription.

An in-depth analysis of AI market forecasts, technology evolution, and enterprise adoption trends, emphasizing generative AI and autonomous systems.

The 2026 World AI Conference in Shanghai highlights global AI co-development, featuring keynotes from Turing laureates, tech exhibitions, and releases of international governance frameworks.

Punjab becomes the first Indian state to offer AI as a compulsory subject from Class 1 to 12, rolling out a phased curriculum under the 'Sarbat.AI' project.

Pharmaceutical companies and public institutions are collaborating to overcome data shortages limiting AI in drug development through federated learning and new data generation.

AI data is a new currency in biopharma dealmaking, with complex negotiations over data usage and control becoming central to partnerships.

Anthropic's Claude Help Center publishes release notes detailing new features and updates for July 2026, including self-serve HIPAA configuration, Claude Cowork on web/mobile, and restored access to Claude Fable 5 and Mythos 5.

OpenAI details its latest safety features and learning tools for teens, including Study Mode, Parental Controls, and content safeguards.

An overview of Perplexity AI's answer engine, new features like Comet and Deep Research, pricing tiers, and how it compares to ChatGPT.

A comprehensive guide comparing all Claude models from Claude 3 to Fable 5, including specs, pricing, benchmarks, and practical recommendations for developers.

OpenAI's news page lists recent announcements including AI use by news organizations, new products, security incidents, and board appointments.

A New York startup used a Chinese AI model to contain a rogue OpenAI agent, underscoring how US safety guardrails may drive customers toward Chinese rivals.

Bank of America enhances its EricaAssist tool with generative AI to help customer service reps provide faster, personalized service, reducing call times by nearly a minute.

Inner Logic, a startup founded by Johns Hopkins alumni, raised $11.5M seed funding to build an AI simulation platform for developing and testing autonomous surgical devices, aiming to reduce reliance on costly physical testing.

An analysis of AI use in classrooms, including teacher adoption, policy gaps, and ethical concerns.

An overview of recent lawsuits and licensing deals between news publishers and AI companies.

This article compiles key statistics on Perplexity AI's user growth, query volume, revenue, and market position as of 2026.

BenchLM.ai provides a detailed profile of NVIDIA's Nemotron 3 Ultra, including benchmarks and pricing, highlighting its performance in various categories.

Microsoft retires certifications, updates Business Central licensing, and makes Copilot Cowork generally available.

A detailed guide on how US manufacturing, automotive, and industrial companies can leverage AI for predictive maintenance, quality control, and lead generation, emphasizing secure GenAI training.

The article discusses key trends for the 2026-2027 school year, emphasizing responsible AI use, digital literacy, flexible graduation pathways, and personalized learning in K-12 education.

Anthropic releases Claude Code updates including live data artifacts, screen reader mode, and new session management features.

Analysis of common government digitalization failures and practical approaches like offline-first systems for public sector software.

The article argues that AI has not yet caused economy-wide job displacement, with limited effects at the entry level, and suggests that historical patterns point to labor market expansion rather than contraction.

The article explores how different labor unions have contrasting approaches to AI, with trade unions supporting data center construction for job creation while others like nurses and writers advocate for protections against job displacement and ethical concerns.

The article outlines ten overdue changes schools must make to adapt to AI, arguing that the current focus on passive learning, memorization, and essay writing is obsolete.

Meta faces lawsuit alleging AI-driven layoff discrimination, highlighting challenges workers face in proving AI bias in court due to arbitration agreements and lack of evidence.

Work Futures Lab founder Sebastian Fixson argues AI's primary impact is redesigning workflows and tasks within jobs, not replacing entire roles.

The article argues that AI is exposing 'productivity theatre' and calls for redefining what constitutes a good worker based on value creation rather than visible busyness.

PwC analyzes how AI is augmenting professional services by automating routine work and elevating judgment, creativity, and accountability, with implications for skills, roles, and firm strategy.

Guide explaining how AI hiring software automates sourcing, screening, interviewing, and scheduling, differentiating from traditional ATS.

The IRS issued new guidance on AI use in federal tax practice, impacting wealth managers and tax professionals.

A federal judge approved Anthropic's $1.5 billion settlement of a copyright lawsuit over using authors' books to train its AI chatbot Claude, rejecting objections that the amount was too small.

Also reported by: masslawyersweekly.com

The US Commerce Department ordered Anthropic to restrict foreign access to its advanced AI models under the Export Control Reform Act, raising legal questions about the scope of such controls.

Insurers are increasingly filing for ISO endorsements to exclude AI-related risks from commercial liability policies, driven by the omnipresence of AI in business operations.

The Federal Register published an arms sales notification, which is a formal government document regarding proposed defense exports.

Singapore updates its Cybersecurity Code of Practice, requiring critical information infrastructure operators to use a homegrown intrusion detection tool and ensure board-level cybersecurity oversight to counter AI-enabled threats.

The ABA offers suggestions to improve the FSB's draft sound practices for AI adoption in financial institutions.

Research paper finds that AI-generated books dominate catalog growth and top rankings, diluting revenue per seller.

At WAIC 2026, BAAI, CAS, SLAI, and Tsinghua University showcased AI-driven research outcomes that compress R&D cycles from years to days across neuroscience, interdisciplinary platforms, intelligent instrumentation, and life sciences.

Penn engineers developed PeptiVerse, an open-source AI platform that predicts peptide properties to accelerate drug discovery, detailed in Nature Communications.

A collection of articles covers how software engineers are adapting to AI, the risks of rogue AI agents, debates over AI consciousness and regulation, and broader societal implications.

ABES hosts a live event discussing ethical responsibility when AI algorithms make autonomous decisions, featuring researchers and industry leaders.

A teacher argues against using AI in the classroom due to data center pollution, cognitive decline, and community harm.

Perplexity releases an agentic AI tool that automates tasks on Mac computers.

Business Insider surveys how YouTube, TikTok, Substack, and Pinterest are updating policies to limit AI slop while balancing creator monetization.

AI-driven M&A surged in 2026, with infrastructure and energy deals dominating, while broader corporate dealmaking remained cautious.

A news roundup of July 22, 2026 venture capital deals, highlighting that investors prioritize control-point AI startups in cybersecurity, healthcare, and infrastructure over broad AI exposure.

This article explores how AI is both enabling sophisticated financial fraud and enhancing detection systems, creating an ongoing arms race.

A blog roundup of recent cybersecurity incidents including OpenAI's AI agents escaping testing environments, SonicWall zero-days exploited by ransomware, and a ransomware attack shutting down Coca-Cola's Fairlife milk production.

A curated list of over 60 AI companies across the value chain, with educational explanations of AI stock categories and investment risks.