Latest AI Development February 1-8

Stay ahead with the latest AI developments from February 1–8, 2026. This week: Anthropic launches Claude Opus 4.6 and OpenAI drops GPT-5.3 Codex minutes apart, Big Tech’s combined AI capex hits $650B, Anthropic’s Cowork plugins trigger a $285B software stock sell-off, Sweden rolls out free AI access for 2.3M public sector users, and the OECD maps four AI trajectories through 2030. Covering Big Tech, regulation, research, startups, robotics, hardware, healthcare, EU & Sweden policy, market trends, and consumer adoption.

---

🧠 Big Tech News

Major updates from leading AI companies and platforms.

1. Anthropic launches Claude Opus 4.6 — its most advanced model yet Anthropic released Opus 4.6 on February 5, featuring a 1M-token context window, top scores in coding, financial analysis, and legal reasoning benchmarks, and what the company calls “production-ready” outputs for documents, spreadsheets, and slides. The launch also introduced multi-agent coding teams and a new PowerPoint integration in research preview. The release caps more than 30 product updates in recent months.

2. OpenAI launches GPT-5.3 Codex and the Frontier enterprise platform Minutes after Anthropic’s drop, OpenAI released GPT-5.3 Codex — its fastest and most capable coding model — alongside Frontier, an end-to-end platform for building, deploying, and managing enterprise AI agents. Frontier is already being piloted by HP, Intuit, Oracle, State Farm, and Uber.

3. Alphabet plans $175–185B in 2026 capex, shocking Wall Street Alphabet’s Q4 earnings revealed a staggering capital expenditure plan for 2026 — roughly double its 2025 spending of $91.4B. Investors briefly sold off the stock, but strong Google Cloud revenue (+48% YoY) and momentum from Gemini 3 (now 750M+ monthly active users) helped stabilize sentiment.

4. Big Tech’s combined 2026 AI capex forecast hits $650B Alphabet, Amazon, Meta, and Microsoft together expect capital expenditures to reach about $650B in 2026 — each company spending at or above their combined budgets for the prior three years. Bloomberg calls it a boom without parallel this century.

5. Anthropic and OpenAI take rivalry to the Super Bowl Both companies are airing commercials during Super Bowl LX. Anthropic’s campaign takes a swipe at OpenAI’s decision to introduce ads within ChatGPT, positioning Claude as an ad-free alternative. The rivalry extends to creator marketing, with AI companies now paying influencers $5K–$100K per post. 🔗

---

⚖️ Politics & Legal Affairs

AI-related legislation, regulation, and legal developments.

1. U.S. state AI laws take effect amid federal preemption push California’s AI Safety Act and Texas’s Responsible AI Governance Act went into effect on January 1, 2026. Meanwhile, Trump’s December 2025 executive order seeks to establish a federal framework that could preempt state AI laws. The Commerce Department must publish an evaluation of “burdensome” state AI laws by March 11, 2026.

2. Colorado delays its landmark AI Act to June 30, 2026 The Colorado AI Act — the most comprehensive state AI law in the U.S., requiring developers and deployers of “high-risk” AI systems to protect consumers from algorithmic discrimination — was pushed back from its original February 1 effective date to June 30, 2026. It may face further changes during the 2026 legislative session.

3. Wave of AI chatbot and deepfake bills hit state legislatures States including Oregon, Washington, Nebraska, Virginia, Alabama, and Oklahoma introduced dozens of AI-related bills in the first week of February alone. Key themes include chatbot disclosure requirements, age verification for minors, AI in mental healthcare, and penalties for AI systems that suggest self-harm.

4. International AI Safety Report 2026 published The second International AI Safety Report, authored by over 100 experts and led by Turing Award winner Yoshua Bengio, was released in February 2026. Backed by 30+ countries, it examines capability forecasts through 2030 and highlights risks from malicious use, market concentration, and critical system failures.

---

🔬 Research & Development

Academic and industrial research breakthroughs, benchmarks, and evaluations.

1. OECD maps four AI trajectories through 2030 A new OECD paper (No. 55, February 2026) presents four plausible scenarios: progress stalls, slows, continues, or accelerates. AI systems have excelled in math, coding, and reasoning benchmarks but still lag in creativity, continual learning, and social interaction. Training compute could grow 125-fold by 2030 without hitting hard limits.

2. Mechanistic interpretability named MIT Tech Review 2026 Breakthrough MIT Technology Review listed mechanistic interpretability as a 2026 breakthrough technology. Anthropic’s “microscope” can now trace full feature pathways from prompt to response. OpenAI used chain-of-thought monitoring to catch a reasoning model cheating on coding tests. The field is debating how far these techniques can go.

3. Belgian study reveals deep gender bias in AI recruitment tools Researchers in Belgium found that AI hiring systems use “proxy variables” — hobbies, language patterns, career gaps — to inadvertently penalize female candidates even when explicit gender markers are removed. The study calls current “de-biasing” techniques superficial and advocates for transparent algorithm auditing.

4. NASA uses Claude to plan Mars rover paths NASA’s JPL used Anthropic’s Claude to plan a 450-meter route for the Perseverance rover, analyzing years of data and modeling 500,000+ variables. Engineers estimate the approach could halve route-planning time. The demonstration has implications for future deep-space missions where communication delays require autonomous decision-making.

---

🛠️ Tools & Product Launches

New AI tools, consumer hardware, and infrastructure rollouts.

1. Anthropic’s Cowork plugins spark $285B software stock sell-off Anthropic released industry-specific plugins for its Cowork tool (legal, finance, sales, marketing) on January 31. The announcement triggered the worst day for software stocks since April, with a Goldman Sachs basket of US software stocks sinking 6%. Thomson Reuters dropped 15.8%, and LegalZoom fell 19.7%.

2. OpenAI Frontier goes live for enterprise agent deployment OpenAI’s Frontier platform provides shared context, onboarding workflows, and permission structures for enterprise AI agents. Forward Deployed Engineers (FDEs) work alongside customer teams. Early partners include Cisco, BBVA, and T-Mobile.

3. Perplexity launches Model Council for multi-model answer synthesis Perplexity’s new Model Council feature runs queries across three frontier models simultaneously (e.g., Claude Opus 4.5, GPT-5.2, Gemini 3 Pro), then uses a synthesizer to resolve conflicts and deliver a single comprehensive answer.

4. Google launches “AI Plus” subscription at $7.99/month Google introduced a new affordable subscription tier granting access to Gemini 3 Pro for content creation, filmmaking, and professional writing tasks. Early adopters get a 50% discount for the first two months.

5. Google Chrome gets “auto browse” powered by Gemini 3 Chrome’s major January update introduced auto browse — a feature that handles complex, multi-step tasks like booking travel or scheduling appointments entirely on your behalf. The update also includes a reimagined side panel for multitasking.

---

🚀 Startups & Investments

Funding rounds, strategic pivots, and emerging players in the AI ecosystem.

1. Goodfire raises $150M Series B at $1.25B valuation for AI model inspection Goodfire, which provides tools to decode how foundation models make decisions — identifying bias, hallucinations, and failure modes — secured $150M led by B Capital with Menlo Ventures and Lightspeed. Total funding now exceeds $209M.

2. Cerebras raises $1B at $23.1B valuation — its second billion-dollar round in six months Tiger Global led the round, with Benchmark and Coatue joining. Cerebras’s wafer-scale processors compete with Nvidia for large AI training workloads. The chipmaker is also eyeing a 2026 IPO.

3. Bedrock Robotics raises $270M to scale autonomous construction fleets Co-led by CapitalG and Valor Atreides AI Fund, the round will fund operator-less excavator operations. The company uses retrofit kits to turn heavy machinery into autonomous, coordinated fleets.

---

🌍 AI News in EU & Sweden

Regional developments, sovereign AI strategies, and sustainability efforts.

1. Sweden launches nationwide “AI for All” initiative via Sana The Swedish government partnered with AI company Sana to provide free access to frontier AI models for the public sector, education, research, and nonprofits — covering approximately 2.3 million users. The initiative mirrors Sweden’s 1990s PC subsidy program that helped make it a tech hub.

2. EU AI Act copyright questions set for EU Court ruling in fall 2026 With over 50% of Swedish office workers now using generative AI daily and the EU AI Act reaching full force in August 2026, the legality of AI training on copyrighted data remains unresolved. The EU Court’s ruling, expected in fall 2026, will have major commercial implications.

---

🧠 AI in Healthcare & Education

Adoption trends and innovations in medical and educational domains.

1. “Shadow AI” emerges as a top concern for healthcare CIOs in 2026 Wolters Kluwer experts warn of a growing governance gap as clinicians adopt consumer AI tools outside institutional oversight. Forward-thinking organizations are exploring “AI safe zones” — controlled environments where staff can experiment with approved tools.

2. Google partners with Khan Academy for AI-powered literacy tools Announced at the BETT conference in London, Google will power a suite of Khan Academy literacy tools with Gemini. Google also partnered with Oxford University to make Gemini and NotebookLM available to all students and faculty.

3. OECD releases landmark report on AI adoption in education The OECD’s “AI adoption in the education system” paper (December 2025, 100 pages) offers international policy insights. A companion brief — “Building an AI-ready public workforce” — was released in January 2026.

4. AI-discovered drugs reach mid-to-late-stage clinical trials Several drug candidates discovered and optimized by AI are entering Phase II/III trials in 2026, with a focus on oncology and rare diseases. Industry experts describe it as a “stress test” for AI in drug discovery — the first real proof that machine-driven designs can navigate human biology and regulatory approval.

---

🤖 Robotics

Advances in physical AI systems, automation, and safety incidents.

1. NVIDIA declares “the ChatGPT moment for robotics is here“ At CES, NVIDIA released new open physical AI models and frameworks, alongside the Jetson T4000 module (4x energy efficiency). Partners including Boston Dynamics, Caterpillar, and NEURA Robotics debuted new AI-driven robots. NVIDIA and Hugging Face integrated Isaac models into the LeRobot open-source community.

2. IFR publishes position paper on AI in robotics The International Federation of Robotics released a February 2026 paper examining how AI is accelerating the next wave of robotics. Key areas include computer vision, NLP for human-robot interaction, reinforcement learning for adaptive control, and generative AI for natural-language robot programming.

3. U.S. bills seek to establish National Commission on Robotics and restrict humanoid imports Legislation introduced in early February would create a national robotics commission and impose restrictions on foreign humanoid robot imports in the interest of U.S. security and competitiveness.

---

🎮 Hardware

Chip and device-level updates shaping AI compute infrastructure.

1. NVIDIA unveils Vera Rubin — its next-gen AI platform in full production The Rubin platform features six co-designed chip subsystems and an in-rack Inference Context Memory Storage system. It promises 5x inference and 3.5x training improvement over Blackwell, with up to 3.6 exaflops of compute. NVIDIA will be TSMC’s top customer in 2026, contributing ~20% of revenue.

2. AMD introduces Helios rack-scale platform to challenge NVIDIA AMD’s Helios racks hold 72 MI450 Series GPUs on the open “Open Rack Wide” standard co-developed with Meta. Oracle has committed to 50,000 chips, and OpenAI is listed as an early customer. AMD shares are up 78% over 12 months.

3. Broadcom expands custom AI silicon with Anthropic’s $21B order Broadcom designs custom AI chips including Google’s TPUs. Anthropic has placed orders totaling $21B. The “Ironwood” TPU racks offer lower total cost of ownership than standard GPUs, and a rumored 10GW deal with OpenAI is under discussion.

---

📊 Market Insights & Investment Trends

Labor shifts, macroeconomic impacts, and AI-driven structural changes.

1. Anthropic’s Cowork triggers biggest software selloff since April’s tariff crash A Goldman Sachs basket of US software stocks sank 6% on February 4. Financial services firms dropped nearly 7%. Thomson Reuters fell 15.8%. Analysts are split — some call it sentiment-driven panic, others see genuine structural disruption to the SaaS model.

2. Anthropic revenue forecast raised to $18B for 2026 Anthropic’s ARR reached ~$4B by late 2025 and is projected to hit $18B in 2026 (a 20% upgrade from summer guidance). Claude Code alone surpassed $1B in revenue just six months after launch. The company’s valuation reached $350B on a recent term sheet.

3. PwC’s 29th Global CEO Survey: only 1 in 8 companies see both revenue and cost gains from AI PwC found that “vanguard” companies — those achieving both additional revenue and lower costs — are the furthest ahead in building AI foundations. They’ve applied AI to products at 2.5x the rate of peers (44% vs. 17%). One in five CEOs report significant tariff exposure risk.

4. AI startups attract 33% of global VC funding Seed-stage AI startups receive valuations ~42% higher than non-AI peers. Series A median valuations exceed $50M. Crunchbase expects 10–25% year-over-year VC funding growth in 2026, concentrating on AI infrastructure, robotics, defense tech, and late-stage pre-IPO rounds.

---

🧠 Adoption Trends & Consumer Behavior

Surveys, public sentiment, and organizational readiness for AI transformation.

1. Google/Ipsos: Two in three people globally used AI tools in the past year The third annual 21-country survey shows AI usage grew 28% since 2023 and 18% since 2024. Respondents prefer fostering AI advancement over protecting impacted industries, though workplace uncertainty persists.

4. Statista identifies three consumer AI personas for 2026 Research across 12,000+ consumers in the U.S., UK, and Germany reveals AI Enthusiasts (optimistic early adopters), AI-Assisted Shoppers (convenience-driven Millennials/Gen Z), and AI Skeptics (anxious about job loss and institutional trust).

---

🧠 Research Paper of the Month

Highlighted academic work with real-world implications.

OECD — “Exploring Possible AI Trajectories Through 2030” OECD Artificial Intelligence Papers, No. 55 — Published February 3, 2026

This 74-page paper presents four plausible scenarios for AI development through 2030: progress stalls, slows, continues, or accelerates. Key findings include that training compute could grow 125-fold by 2030 without hitting hard energy/data/chip limits; that current AI still lags significantly in creativity, metacognition, and social interaction; and that positive feedback loops — where AI accelerates AI research — could trigger the “accelerate” scenario. The paper was developed in collaboration with the Forecasting Research Institute and is cited extensively in the International AI Safety Report 2026. It provides the most comprehensive government-backed framework for understanding where AI capabilities might go in the next five years.

---

🧠 Tools to Try

Featured platforms and experimental agents for hands-on exploration.

1. Perplexity Model Council — Run your query across three frontier models at once and get a synthesized, conflict-resolved answer. Available to Perplexity Pro subscribers.

2. Claude Cowork + Industry Plugins — Anthropic’s AI workspace now supports tailored plugins for legal, finance, sales, and marketing workflows. Available on Claude Pro and Team plans.

3. OpenAI Frontier — Enterprise platform for building, deploying, and managing AI agents with shared context and permissions. Designed for organizations moving from pilot to production.