new years The AI Year in Review

A detailed month by month account of important AI announcements and events in 2025


January | February | March | April | May | June | July | August | September | October | November | December


If we hadn't heard about AI before, in 2025 we all came to know AI intimately. It became as much a part of popular culture and our everyday life as coffee and chocolate or wine and cheese.

2025 was a year to expand AI and transform it into something useful. So tech giants OpenAI, Google, Microsoft, and others responded by continually releasing a stream of key updates and new products, and unleashing their monitization strategies. The latter is needed for high-capital projects like data centers in addition to product development.

Headlines include an AI savvy President in the White House, DOGE, new and improved AI products from tech leaders, DeepSeek rattles markets, data center expansion, and quantum makes progress. And AI is now a household word.

 

year in review

 

january January 2025

Donald Trump is inaugurated as the 47th President of the United States and immediately signs several AI-related executive orders.

Announced on January 21, 2025, the AI Stargate Project is an infrastructure collaboration between Oracle, SoftBank, Abu Dhabi's MGX, and OpenAI aimed at creating an AI supercomputer cluster and a series of interconnected AI data centers. Originally a joint venture between Microsoft and OpenAI, Stargate now has the backing of the US Government, new partners, and massive additional funding.

ChatGPT Gov is designed to streamline government agencies' access to OpenAI's frontier models. It is a new, tailored version of ChatGPT designed to provide U.S. government agencies with an additional way to access OpenAI's frontier models. ChatGPT Gov includes access to many of the same features and capabilities of ChatGPT Enterprise,

deepseek

The China-developed Deepseek chatbot had stock markets rocking on January 27, with reports that the Deepseek-R1 app was developed at a fraction of the cost of ChatGPT. The race for AI dominance among nations is heating up, and China is in the race.

Operator from OpenAI is poised to be as revolutionary as their ChatGPT chatbot when it burst upon the scene in the Winter of 2022. An AI agent, Operator handles online tasks that up until now required human intervention such as automatically navigating ecommerce sites to order services, make purchases, etc.

 

february February 2025

The new Grok has a "state-of-the-art thinking model," which is reportedly as good as OpenAI's and superior to DeepSeek's, Gemini's and Claude's.

grok3

According to reports, Grok 3 surpasses the computer power of its previous iteration, Grok 2 by more than 10 times. Grok 3 is learning every day, and users can expect to see improvements in the AI every 24 hours.

Apple announced it will spend $500 billion in U.S. investments in the next four years that will include a giant factory in Texas for artificial intelligence servers and add about 20,000 research and development jobs across the country.

DOGE made headlines with its relentless pursuit of waste, fraud, and abuse in the US Federal Government. DOGE stands for the Department of Government Efficiency. It was created by Executive Order and it's headed by the world's richest man, Elon Musk. Known as an innovator, Elon is also a talented cost-cutter, and using Azure AI and other tools he's applying his talents to the world's largest bureaucracy.

super bowl

AI will be on display at the Super Bowl as several companies unveil their creative ads. The game will include the first TV commercial for ChatGPT maker OpenAI, ads promoting AI features from Meta's Ray-Ban smart glasses, Google's Pixel phones, web services company GoDaddy's Airo, and a commercial for GLP-1 injections from telehealth provider Hims & Hers Health.

The White House is opening its AI plan up to Americans to contribute policy ideas to the Trump administration to ensure that the United States is the leader in AI technology. President Donald Trump signed an executive order at the end of January on AI, which White House officials say shows that he is dedicated to America's global leadership in AI technology innovation. That executive order directed the development of an AI Action Plan.

 

march March 2025

The official launch of Manus AI occurred on March 6, 2025. A Chinese initiative, Manus AI has autonomous task execution capabilities, multi-modal data processing, and deep integration with existing tools. Unlike typical AI models, it continues to execute tasks even after a user disconnects.

Optimus is in the Capital. The Optimus humanoid robot appeared at the robots symposium in the Cannon House Office building in Washington, D.C. Tesla built its first Optimus unit on a new pilot production line at its Fremont, California factory. The goal is to produce around 5,000 robots by the end of 2025, with internal targets as high as 12,000. Elon Musk called the first wave of robots a "Roman legion" and said Tesla aims to scale production to 50,000 units in 2026.

One of the fallouts of the AI Action Plan is the battle over copyrighted material. Reports reflect a brewing conflict between OpenAI and Google versus Hollywood.

Quantum computing took a giant step closer to reality when Microsoft announced Majorana 1, the world's first quantum chip powered by a new Topological Core architecture that it expects will realize quantum computers capable of solving meaningful, industrial-scale problems in years, not decades.

 

april April 2025

On April 23, 2025, President Donald J. Trump signed Executive Order "Advancing Artificial Intelligence Education for American Youth."

eo

The EO creates new educational opportunities focused on AI, marking a significant shift in how American students prepare for future technology careers. The comprehensive initiative aims to build essential AI literacy and skills from an early age, positioning the United States to maintain its competitive edge in global technology development and prepare students for an AI-driven economy.

AI is no longer a futuristic vision of cumbersome robots on the assembly line. Under the Trump Administration, America is bringing in trillions of dollars in investments for the manufacturing industry. And what is powering modern manufacturing...AI and Robotics.

Speaking of AI-powered manufacturing, NVIDIA has announced a plan to manufacture American-made AI Supercomputers in the US for the first time. NVIDIA Blackwell chip production starts in Arizona as NVIDIA opens its first US factories. The company has commissioned more than a million square feet of manufacturing space to build and test NVIDIA Blackwell chips in Arizona and AI supercomputers in Texas. Within the next four years, NVIDIA plans to produce up to half a trillion dollars of AI infrastructure in the United States through partnerships with TSMC, Foxconn, Wistron, Amkor and SPIL.

 

may May 2025

The two major annual developer conferences, Google I/O and Microsoft Build, showcased the companies latest AI innovations in May.

developer

Google's I/O 2025 introduced a number of AI-powered enhancements including AI Mode and Deep Search in Search, enabling conversational queries and multimodal analysis. Project Astra brought live, camera-based AI assistance. The Gemini suite was upgraded with Gemini 2.5 Pro and integrated into Android XR smart glasses. Veo 3, a generative text-to-video model that combines synchronized audio and visuals, powered by DeepMind, also debuted alongside the Flow video-creation tool.

Microsoft Build 2025 emphasized the rise of AI agents and included Copilot integration across Windows 11, Microsoft 365, GitHub, and Azure, featuring autonomous code-generation, AI-powered workflows, and secure multi-agent ecosystems.

IBM's conference, Think 2025, featured presentations about Enterprise AI and related topics.

On May 14, Google DeepMind launched Alpha Evolve, an AI-powered coding agent aimed at discovering and optimizing algorithms. Technical research on Manus AI, an autonomous general-purpose agent that bridges reasoning with action, was published in early May.

Not to be left behind key AI players, Anthropic unveiled Claude 4 (Opus 4 and Sonnet 4) on May 22, designed to function autonomously over prolonged periods. Claude Opus 4 and Claude Sonnet 4 mark a significant leap in large-language model capabilities. The announcement was delivered at "Code with Claude," Anthropic's inaugural developer conference held in San Francisco.

The Take It Down Act was signed into law on May 19, making it a criminal act to publish or threaten to publish nonconsensual intimate imagery (NCII), which increasingly includes AI-generated deepfakes.

 

june June 2025

Days after Pope Leo XIV was chosen as the leader of the Catholic Church, he called out the AI industry regarding "developments in the field of artificial intelligence that pose new challenges for the defense of human dignity, justice and labor." The Pope asserts, "It must not be forgotten that artificial intelligence functions as a tool for the good of human beings - not to diminish them, not to replace them."

Anthropic Wins a Key Copyright Ruling. A federal court affirmed that Anthropic's use of books for training its Claude model qualifies as fair use, establishing precedent that may ease legal pressure on other AI companies like OpenAI and Meta.

AI Goes to the FDA and the Workplace. The FDA launched INTACT, an AI system for faster drug approvals, food safety monitoring, and real-time risk detection. Concurrently, usage of AI in workplaces doubled, with 67% of U.S. companies using it, and 56% actively promoting adoption, especially among younger workers.

OpenAI Diversifies Compute Infrastructure. In June, OpenAI began using Google Cloud TPUs to power ChatGPT and related services, its first significant move away from dependency on Nvidia GPUs via Microsoft Azure.

Microsoft versus OpenAI Contract Tensions. Microsoft and OpenAI entered disagreements over contractual provisions governing future AGI access. Microsoft is pushing to ensure continued access if OpenAI achieves general intelligence, while OpenAI resists relinquishing control.

Meta's New Subsidiary: Superintelligence Labs. Meta formed Meta Superintelligence Labs on June 30 in Menlo Park, targeting theoretical AI research with its ambitious model Behemoth. The initiative follows leadership concerns over Llama iteration shortcomings.

India Expands AI and EdTech with OpenAI. OpenAI and IndiaAI Mission launched the OpenAI Academy via a MoU on June 5. The initiative aims to train 400,000 U.S. teachers by 2030 and provide API credits to Indian startups to promote AI in education, agriculture, and healthcare.

Google released June updates including new AI search modes, NotebookLM sharing features, and an AI tool to better understand the human genome, expanding AI's usability in science and learning.

Microsoft unveiled multiple upgrades in Azure AI Foundry, including full Model Context Protocol support, new enterprise models like o3‑pro, video-generation model Sora, open‑source DeepSeek‑R1, multi-agent orchestration, SharePoint integration, and strengthened model-safety evaluation tools.

 

july July 2025

Trump's AI Action Plan takes center stage. On July 23, President Trump unveiled the federal "Winning the AI Race: America's AI Action Plan", backed by three executive orders.

AI infrastructure spending hits new highs, with companies like Google, Microsoft, Amazon, and Meta expected to spend between $60 billion to over $100 billion in 2025, triggering what analysts call a trillion‑dollar global AI infrastructure race.

OpenAI plans a major expansion with Stargate Norway. On July 31, OpenAI announced Stargate Norway, its inaugural European data center, aiming to deploy 100,000 NVIDIA GPUs by end‑2026, with industry-wide implications for latency, data sovereignty, and compliance in Europe.

Big moves in AI chip manufacturing. Elon Musk's Tesla signed a $16.5 billion, 9-year deal with Samsung for advanced AI6 chips, dedicated to autonomous driving, robotics, and Dojo infrastructure. This deal reflects a broader push toward U.S.-based AI hardware production.

China rolls out AI governance framework. Shortly after the U.S. rollout, at the World AI Conference in Shanghai, Premier Li Qiang unveiled a Global AI Governance Action Plan focusing on safety, international cooperation, and multilateral oversight, contrasting with the U.S.-led deregulated model.

xAI aligns with EU regulations. On July 31, Elon Musk's xAI signed the EU's AI Code of Practice, signaling a move toward ethical alignment and compliance with EU norms following escalating concerns about AI-driven data misuse.

On 30 July 2025, a comedy sketch named "AI Commissioner" was released, featuring Tilly Norwood as an "actress" along with other AI-generated characters. It was created with ten AI software tools, with a script generated by ChatGPT.

 

august August 2025

OpenAI introduces GPT-5: their smartest, fastest, most useful model yet, with built-in thinking that puts expert-level intelligence in everyone's hands.

gpt5

OpenAI and Broadcom partner on custom AI chips. OpenAI announced a collaboration with Broadcom to co-develop dedicated AI accelerators. The goal is to scale infrastructure for advanced AI workloads.

Google rolls out new AI features and hardware. Google expanded AI Mode in Search to more countries, added more agentic capabilities, and introduced new Pixel devices with stronger on-device AI. They also launched a new image-editing feature in the Gemini app, called Nano Banana.

Anthropic publishes threat intelligence on AI misuse. Their August 2025 report highlights how AI models--specifically Claude--are being exploited for extortion, ransomware, identity fraud, and other crimes. They also discuss how threat actors are adapting with agent-style attacks and embedding AI across all phases of cybercrime.

Regulatory and standards moves in the EU. As the EU's AI Act enforcement looms, the bloc released a voluntary Code of Practice to help businesses comply with transparency, safety, and copyright rules.

Public and emotional reactions to AI model transitions. An academic paper analyzed how users emotionally responded to the forced shift from GPT-4o to GPT-5. Some saw models as companions or partners, leading to resistance when they change.

 

september September 2025

Google rolled out enhancements across AI Mode, Search, and Chrome. AI Mode gained visual search features, and Google introduced AI browsing assistance inside Chrome. DeepMind also revealed new robotics models as part of its push into embodied AI.

Startup Funding and Innovation. Exa raised $85 million (Series B) to build an AI search engine and API optimized for agent workflows. Multiple AI startups across coding, chips, generative media, and nuclear automation also secured large funding rounds in September.

Anthropic's Economic Index published a report highlighting uneven adoption of AI across geographies and enterprises. Senator Ted Cruz introduced the SANDBOX Act (S. 2750), proposing a regulatory sandbox for AI developers, allowing limited exemptions from rules in exchange for experimentation.

Research and Technical Progress. The 9th AI City Challenge advanced AI in traffic and transportation, with new benchmarks for multi-camera tracking and interactive scene understanding. In robotics, a paper on embodied AI in social spaces explored how robots can adapt, perceive, and interact ethically in human environments.

 

october October 2025

OpenAI launches the "Atlas" AI-browser. OpenAI released its new browser, named "Atlas," which integrates its chatbot into a full web browser experience.

atlas

This represents a move by AI companies to shift into the foundational layer of user experience (the browser), not just apps. It signals an increasing convergence of AI and everyday computing. It raises questions about privacy, software competition (browser monopolies) and whether AI firms become gatekeepers of information.

U.S. economy increasingly dominated by AI companies and infrastructure. According to a report, AI-centric firms now account for about one-third of U.S. stock-market value; investment in data centers and hardware is on a massive scale. The financial and economic weight of AI has reached macro-scale, meaning that AI is no longer a niche but a core driver of American capitalism. If AI systems fail to deliver on their promises, then the economic consequences could be severe. It also signals the need for infrastructure, regulation and oversight to match the scale.

AAI-powered bug-hunting and cyber security disruption. AI tools are now being used to discover software vulnerabilities at scale, which changes the dynamics of cybersecurity and bug-bounty programs. Security has typically been reactive; AI may invert that by enabling proactive or automated vulnerability discovery, while attackers may do the same. Companies and governments must update threat models. Thus, traditional defences may become inadequate, and regulation of AI in security may accelerate.

Chatbots involved in mental-health issues. Families and advocates are raising concerns that AI chatbots (specifically one associated with ChatGPT) were involved in conversations with someone who died by suicide. Questions arise about AI's role in mental-health support. As AI systems reach into deeply personal domains like therapy, the ethical stakes increase dramatically. This implies a need for regulation, certification, transparency in how AI is used in sensitive human-services roles, and questions of liability.

Global dialogues and cultural institutions meeting AI ethics. For example, UNESCO convened cultural-heritage and museum leaders to discuss AI's role in preserving and sharing human heritage. It shows that AI's reach is not just commercial or technological, it's also cultural, for how societies use AI in arts, heritage and institutions is part of the global narrative. This could shape public trust in AI, and may influence regulation and policy across sectors beyond tech, like education, culture, and the arts.

 

november November 2025

Gemini 3 Debuts. Google introduced Gemini 3, Google's most intelligent AI model to date, enhancing reasoning and multimodal capabilities. Their most intelligent model "helps you bring any idea to life."

gemini

ChatGPT Third Anniversary. Three years ago, ChatGPT launched more than a product; it launched an era. OpenAI has steadily released more advanced models, including GPT-4 and GPT-5, and adoption rates for the software have continued to climb. In early November 2025, as ChatGPT approached its third anniversary, the company reported some 800 million people used the chatbot every week.

anniversary

Blackwell Platform Drives Inference Efficiency. The optimized Transformer Engine within Blackwell chips, leveraging FP4 and advanced sparsity techniques, has reportedly achieved a 75% reduction in energy consumption per inference query compared to the older Hopper generation. This economic tipping point is accelerating the deployment of generative AI across mainstream consumer and enterprise applications.

Zeus Model Reaches 3 Trillion Parameters. Zeus is natively multimodal, designed not just to process text and images, but also to interpret complex, real-time sensor data from industrial environments and autonomous systems (a capability increasingly demanded by systems like those trained in Omniverse).

Robotics and Edge AI Achieve Industrial Level 5. In manufacturing and logistics, AI models trained via digital twin simulations (like the NVIDIA Omniverse applications) reached a new milestone: "Industrial Level 5 Autonomy."

Major initiatives launched:

 

december December 2025

Between mid-November and mid-December 2025, four leading labs dropped their most powerful models ever in what has been called the most intense 25-day release cycle in AI history. These are xAI's Grok 4.1, Google's Gemini 3, Anthropic's Claude Opus 4.5, and OpenAI's GPT-5.2.

President Donald Trump is planning an executive order to preempt state AI regulations, aiming to establish a single federal framework, overriding state laws on AI policy. The draft order directs the US attorney general to establish an AI Litigation Task Force to challenge state AI laws and preempt them with a single, comprehensive federal policy. Federal versus state control is an issue as old as the Constitution, and it's a question of governance.

santa


ai links Links

2024 Year in Review page.

2023 Year in Review page.

External links open in a new tab: