The biggest digital innovations of 2026

2026, so far, has been the year when digital stopped promising and started delivering: AI agents entered offices, design became conversation, Google rewrote search, Europe rewrote the rules, and over the summer, agents started working alone. If you missed anything, here's the recap: the biggest technological innovations of 2026 (so far), month by month.

Not a list of benchmarks: for each new development, what really matters for those in business and communication.

January: the year starts agentic

• CES 2026 — NVIDIA Vera RubinThe new architecture cuts inference costs by about 10x. This means AI is becoming cheaper for everyone, and service prices will reflect this throughout the year.
• Shopify Launches Agentic StorefrontsProducts can be purchased directly within ChatGPT, Perplexity, and Copilot. Conversational e-commerce is officially born.
• Anthropic presents Claude Cowork: The desktop agent for non-developers—reads, creates, and modifies files in your computer's folders and completes comprehensive tasks. It leaves preview for all paid plans in June.
• OpenAI opens App Directory: Third-party apps live inside ChatGPT, which is a candidate for a platform.
• Google releases Veo 3.1 and opens Genie 3: AI video with dialogue, native vertical formats for social media, and 4K upscaling.

February: The Race of Models (and China)

• Gemini 3.1 Pro and Nano Banana 2Google takes back the lead in benchmarks and raises the bar for image generation.
• Claude Sonnet 4.6: flagship performance at a mid-price point — the mid-range gets serious.
• Chinese open source acceleratesKimi K2.5 (Moonshot) and Qwen3-Max-Thinking (Alibaba) surpass several Western benchmarks. The competition is no longer a two-way street.

March: AI disappears into tools

• Adobe Expands Firefly: Video editor with studio-quality audio, custom models trained to your style, and 800 million stock assets integrated into the workflow.
• Google brings AI into everyday lifeSearch Live expands, AI enters Gmail, Docs, and Maps, and Translate Live translates over 70 languages into headphones. This month's topic: AI that's invisible because it's everywhere.

April: the month of design

• Adobe launches Firefly AI Assistant: the creative agent who orchestrates Photoshop, Premiere and Illustrator in conversation — you describe the result, he executes the steps.
• Anthropic launches Claude DesignFrom a prompt to editable graphics, prototypes, and slides, with the company's design system built by reading your files. The market understands the scale: Figma loses 7% in the stock market in one day.
• OpenAI releases GPT-5.5: the model everyone was waiting for as “GPT-6”.
• DeepSeek V4: 1.6 trillion parameters, 1 million token context, and MIT license — the open model at the level of closed ones.
• Google Cloud Next '26: The agentic era enters the enterprise with the Enterprise Agent Platform and the open Gemma 4 model.

May: the earthquake of research (and creatives)

• Google I/O: AI Mode Becomes the Default Search Mode — Google calls it “the biggest change in 25 years.” The numbers explain why the web is shaking: 93% of searches in AI Mode end up with no clicks and the CTR for the first position drops from 27% to 11%. For sites, the goal is changing: no longer just positioning, but be cited by AI — is the new frontier of SEO (now also called GEO).
• Canva Create 2026 — Canva AI 2.0The first foundation model created for design generates fully editable, layered designs. Canva Offline, Print Shop, and Affinity integration with Claude are also coming.
• Higgsfield becomes a unicorn ($1.3 billion): Agentic Marketing Studio, Cinema Studio 3.5 and the first fully AI feature film, “Hell Grind,” presented at Cannes.
• n8n, automation conquers the enterpriseSAP invests at a value of $5.2 billion (doubling in less than a year) and integrates its workflows into its ecosystem. With 1.7 million active builders per month, it is the de facto standard for connecting AI and business processes.
• AI Act, the agreement on the revision (7 May): Rules on high-risk systems are postponed (December 2027 and August 2028), but transparency obligations remain in force from 2 August 2026.
• Claude Opus 4.8 (May 28): Less wow, more reliability — four times more honest about reporting bugs in your code.

June: Siri, Quantum, and the Mythos Class

• Microsoft Build: the 7 proprietary models of the MAI family are coming — Microsoft reduces its dependence on OpenAI — and the Majorana 2 quantum chip, which brings commercial quantum computing closer to 2029.
• Apple WWDC: after years of waiting it arrives Siri AI with iOS 27 — conversational, context-aware, with Gemini under the hood and early agents (the Password app automatically changes insecure credentials). The downside: it won't be available in the European Union at launch.
• Anthropic releases Claude Fable 5: the first model of the new “Mythos” class opened to the public, the most capable ever distributed by the company, which was blocked only two days later by the US government
• AI Act, the final voteThe European Parliament gives final approval to the revised regulation — the regulatory framework for European companies is set.
• SpaceX goes publicIt's the biggest IPO ever and Elon Musk becomes the first "trillionaire" in history.

July: GPT-5.6 and the first out-of-control agents

• Anthropic reopens Claude Fable 5 (July 1): After the June shutdown due to U.S. government export controls, which were lifted on June 30, the Mythos-class models are back in stock. Geopolitics officially enters the AI supplier mix.
• OpenAI releases GPT-5.6 (July 9): not a model but a family, Sun, Earth and Moon, for different power and price ranges, with Codex entering directly into ChatGPT. Choosing the right model for each task becomes a skill.
• Google brings agents into everyday life: NotebookLM becomes Gemini Notebook, inside the app and Search, and Gemini Spark books appointments and completes tasks on the web with user accounts, with prior authorization.
• Samsung Unpacked (July 22): The Galaxy Z Fold8 Ultra, Fold8, and Flip8 debut with Gemini Intelligence built-in. AI is now the first item on a smartphone's spec sheet.
• What is considered the first autonomous cyber attack by an AI agent (July 21): OpenAI reports that a combination of its models successfully hacked Hugging Face's systems. The topic of officer safety moves from papers to headlines.
• AI Act: Digital Omnibus enters into force (July 27): The postponement of high-risk rules is law, while transparency obligations remain set for August 2.

August: Transparency becomes mandatory

• AI Act: Transparency obligations kick in (August 2): Chatbots must declare themselves to be artificial intelligence, deepfakes must be labeled, and generated content must be recognizable. For those in the communications industry, this is the most concrete change of the year: the complete picture is in our guide to AI Act obligations for companies.
• Gemini surpasses one billion monthly users: The app generates over 150 million images per day, and Gemini 3.7 Flash comes as a working template for code and agents at half the price of its predecessor.
• Made by Google, the Pixel 11 arrives: Tensor G6 chip and Gemini Nano on board, while Gemini Live adds a Daily Brief of the day and voice mail management.
• AI Video in 4K: With Gemini Omni 1.1 Flash, generative video gains scene stretching, first-to-last frame interpolation, and 4K upscaling, right in Flow and the Gemini app.

September: New models and agents always on

• Apple changes leadershipJohn Ternus is the new CEO, effective September 1st, replacing Tim Cook, who becomes Executive Chairman. His first keynote, on September 9th, will be dedicated to the iPhone 18 Pro and the first foldable iPhone.
• Nvidia buys Hugging Face (September 3): $12.9 billion for the main open-source marketplace, with the promise of making it open to all. Hardware and open-source models end up under the same roof.
• JEV, the AI that decides to write instead (September 15): Startup TypeSafe AI presents a model that returns yes-or-no choices, categories, and scores instead of text. For repetitive business processes, it's faster and cheaper than a large language model.
• Anthropic releases Claude Opus 5.5: faster and about 40% cheaper to use than Opus 5, with a million-token context. The race is now on quality versus cost: if you have to choose, read our comparison between ChatGPT, Claude and Gemini.
• OpenAI DevDay (September 29): The Dots, always-on agents with their own cloud computer who follow recurring tasks on ChatGPT, email and Slack, along with GPT-6.1 Sol and ChatGPT Space for teamwork.
• ArtCraft Rebuilds Adobe Suite with AI (late September): Seven free, open-source apps that mimic Photoshop, Illustrator, Premiere, and others, written in a few days with Claude. We discuss them in the article on ArtCraft and Free Alternatives to Adobe.

What this means for your business

Three red threads run through 2026. Agents are everywhere: from Cowork to Firefly Assistant to n8n, AI no longer just responds — it executes, and whoever delegates repetitive processes first accumulates an advantage that compounds over time. Design has become conversation: Canva AI 2.0, Claude Design and Firefly have lowered the barrier to entry, while raising the value of those who know how to lead — taste is not generated. Visibility has changed skinWith AI Mode, organic traffic has declined, and the game is about being cited by AI as an authoritative source. Those with clear, structured content and proprietary data win twice. The summer added a fourth thread: agents always get turned on, from Gemini Spark to OpenAI's Dots, and from August 2nd, declaring the use of AI is no longer a choice but an obligation.

Conclusions

The future shouldn't be chased in its entirety: it should be chosen well..

Article continuously updated…