Through 2023, generative AI moved from standalone novelty products into the software people already used daily. Adobe launched Firefly, its image-generation model, in March 2023 and integrated it directly into Photoshop's 'Generative Fill' tool by May, letting users extend or edit photos with AI-generated content inside a workflow with hundreds of millions of existing users rather than a separate app. Microsoft rolled out Copilot across Word, Excel, and Outlook starting in March 2023, embedding chat-based drafting and summarization into its core productivity suite.
The mechanism was distributional rather than purely technical: rather than requiring users to visit a new website or learn a new interface, generative capability arrived as an upgrade inside tools with existing habits and muscle memory, dramatically lowering the adoption barrier compared to the ChatGPT model of a standalone destination. GitHub Copilot, already available since 2021, saw usage surge alongside broader AI awareness, and its parent Microsoft reported it was writing a large share of code in supported languages within completed pull requests by developers using it.
Legal infrastructure lagged the technology sharply. Getty Images sued Stability AI in February 2023 alleging its Stable Diffusion model was trained on millions of copyrighted Getty photos without license, and comedian Sarah Silverman joined a class-action suit against OpenAI and Meta over training data later that year. The U.S. Copyright Office ruled in February 2023, in the 'Zarya of the Dawn' case, that AI-generated images within a human-authored comic could not themselves receive copyright protection, establishing an early — and still contested — line between human and machine authorship.
Creative professionals split sharply on outcomes: stock photography and illustration freelancers reported falling demand for commodity work that generative tools could now approximate, while designers who integrated AI into faster iteration workflows reported productivity gains. Publishers and news organizations, watching AI models trained partly on their content power products that competed with their traffic, moved toward licensing deals — Axel Springer and the Associated Press struck agreements with OpenAI in 2023 — rather than relying solely on litigation.
Coverage through 2023 focused heavily on generative AI's most visually striking outputs — AI-generated art winning contests, viral fake images of public figures — over the more consequential but less photogenic integration happening inside enterprise software contracts, where AI features were being bundled into subscription tiers that quietly changed the economics of Adobe, Microsoft, and Salesforce's core businesses.
By the end of 2023, essentially every major productivity and creative software vendor had shipped or announced a generative AI feature, and Google rebranded and relaunched its chatbot as Gemini in early 2024, integrating it more deeply into Search, Android, and Workspace. The New York Times filed suit against OpenAI and Microsoft in December 2023 over alleged use of its articles in training data, a case that, as it proceeded through the following years, tested whether 'fair use' doctrine — built for a pre-generative-AI world — could accommodate models trained on the entire searchable internet.
Image, code, and video generators moved from demos to Adobe, Microsoft, Google, and startup toolchains. Creative-labor markets felt pricing pressure; 'AI-assisted' became a résumé line and a disclosure fight. Capital expenditure on GPUs reordered Big Tech budgets.
Evaluation benchmarks saturated; product moats shifted toward data, distribution, and reliability. Regulators began drafting rules mid-explosion. Mainstreaming means the technology is no longer optional context for knowledge work — it is ambient infrastructure with unsettled norms.
Hollywood strikes put AI clauses on picket signs; classrooms put detectors on essays. The culture fight is really a labor and authenticity fight. Mainstream means the argument moved from labs into contracts, classrooms, and courts.
Procurement teams now ask vendors for eval harnesses and data-retention terms the way they once asked for uptime SLAs. The technology crossed from novelty budget into operating budget — and operating budgets demand accountability paperwork. That residue still structures decisions.
The present inheritance is that 'AI feature' has become a standard line item in enterprise software pricing and product roadmaps rather than a differentiator, and copyright litigation from the 2023 wave continues to shape licensing norms years later, with publishers now routinely negotiating AI-training rights as a distinct revenue line separate from advertising or subscriptions — a category that did not meaningfully exist before that year.
Century Signals note: Major product launches 2023–2024; earnings commentary on AI capex; labor-market and copyright reporting. Editorial judgment about what still structures the present — not a comprehensive history.