20 May 2026: Google Rebuilds Search Around AI, Stability Opens Up Audio, and Musk Loses to OpenAI

Google’s annual developer conference delivered the most consequential AI search overhaul in two decades, replacing the familiar keyword bar with an interface that understands images, voice, and entire browser tabs. Stability AI has opened its music generation models to the world. And a California jury ended Elon Musk’s $134 billion lawsuit against OpenAI in under two hours.

Google has redesigned its search box for the first time in 25 years, replacing the familiar keyword bar with an AI-native interface that understands text, images, voice, video, and even open browser tabs. The announcement came at Google I/O 2026 and marks a fundamental shift: Google is no longer just a tool you query, but a system that can act on your behalf. The new AI Mode can ingest entire documents or live camera feeds as inputs, while a new Search Agents feature lets users build persistent mini-apps that continuously crawl the web and pull in data from Gmail, Calendar, and Photos. For UK consumers and businesses that rely on Google Search as their starting point for research, procurement, or marketing, this is a foundational change in how that tool works. If you want to understand what these persistent helpers can realistically do, our explainer on what AI agents can actually do today covers the essentials.

Google also announced Gmail Live, a voice-driven assistant built directly into your inbox that lets you speak questions about travel bookings, schedules, or older conversations and receive answers drawn from your email history. Gmail Live will initially roll out to subscribers of the new Google AI Ultra plan, priced at $100 per month. Voice-first AI built into a productivity suite represents a significant practical step for small businesses managing high email volume. If you are currently weighing whether an AI subscription is worth the cost, our comparison of free AI tools versus paid AI tools sets out what you actually get at each price point.

Alongside the search changes, Google released Gemini 3.5, a new series of models that combines frontier intelligence with agentic action capabilities. The headline model, Gemini 3.5 Flash, delivers four times the output token generation speed of competing frontier models, according to Google, making it well-suited for developers building applications that need fast, high-volume responses rather than raw accuracy alone. Google is also releasing Gemini Omni, a multimodal model capable of conversational video editing and creative generation across text, image, audio, and video inputs. All new models are available through the Gemini API. The pace of Google’s announcements at this year’s I/O suggests the company is moving decisively to close the gap with OpenAI and Anthropic on both capability and developer tooling, at a moment when enterprise AI spending is accelerating.

Stability AI has released Stable Audio 3.0, a family of four music and sound generation models, with open-weight versions that anyone can download and run for free. The lineup runs from a 459 million parameter small model, capable of generating two minutes of audio and suited for on-device use on a smartphone or laptop, up to a 2.7 billion parameter large model that can generate full compositions of up to six minutes and twenty seconds. The small and medium models are being released with open weights, meaning developers and creators can download, modify, and deploy them without a commercial licence. The larger models significantly outperform Stable Audio 2.0 in their ability to maintain melodic structure and coherence over extended tracks. For podcasters, indie game developers, and content creators who currently pay for music licensing, this is a meaningful practical alternative worth testing right now.

A California jury unanimously rejected Elon Musk’s lawsuit against OpenAI and CEO Sam Altman on Monday, ending a two-year legal battle in under 90 minutes of deliberation. Musk had sued in February 2024, claiming Altman and co-founder Greg Brockman had “stolen a charity” by shifting OpenAI from a non-profit to a structure with a for-profit arm, and he sought up to $134 billion in damages alongside Altman’s removal as CEO. The jury found Musk had exceeded the applicable statute of limitations when he filed his case. Judge Yvonne Gonzalez Rogers confirmed the verdict, describing the supporting evidence as substantial. Musk’s legal team has announced plans to appeal. For OpenAI, the verdict removes a significant legal overhang as the company continues its pending for-profit restructuring.

A federal appeals court in Washington heard arguments on Tuesday in Anthropic’s lawsuit against the US Department of Defense, which designated the Claude maker a supply chain risk to defence contractors in March. The designation requires military contractors to replace Anthropic’s models in their systems. Anthropic argues the DOD is misusing a narrow procurement label to gain leverage in what is fundamentally a contract dispute, while the Pentagon’s legal team said the designation was needed to notify the agency of risk as quickly as possible. In a notable signal for Anthropic, Judge Henderson said the DOD’s actions appeared to be a “spectacular overreach,” adding that she saw no evidence supporting the supply chain risk finding. No ruling has been issued. UK-based organisations operating across NATO-aligned supply chains may want to monitor this case: a ruling in Anthropic’s favour would be the first time a US court has directly constrained how the government can restrict commercial AI models from defence procurement frameworks.

Here is everything else worth knowing from today’s AI news.

• OpenAI expands Education for Countries: OpenAI is advancing its multi-year programme to bring AI tools to schools globally, adding new partnerships, teacher training, and classroom tools. OpenAI
• NanoClaw creator turns down $20M buyout, raises $12M seed instead: NanoCo, the company behind AI agent platform NanoClaw, rejected an acquisition offer after a viral launch and closed a $12 million seed round. TechCrunch
• Mistral AI acquires Emmi AI: European AI lab Mistral has acquired Emmi AI in a move that extends its product and tooling capabilities. Emmi AI
• Ocean raises $28M to fight AI-generated phishing: The agentic email security startup claims its AI analyses the full context of every incoming email to detect fraud and impersonation attempts. TechCrunch
• Figma adds an AI assistant to its collaborative canvas: Figma’s new AI assistant is being rolled out first to Figma Design users and will assist with design tasks directly within the canvas. TechCrunch
• Anthropic tops CNBC Disruptor 50 list: Anthropic has displaced OpenAI as the number one company in CNBC’s 2026 ranking, with CEO Dario Amodei reporting 80 times revenue growth in the first quarter. CNBC
• Google launches AI Ultra subscription at $100 per month: The new tier unlocks Gmail Live, Gemini Pro access, and expanded storage, sitting above the existing Google AI Plus and Pro plans. Google
• AI search startups seeing explosive investment growth: AI-native search has become one of the most actively funded areas in consumer AI, with multiple new entrants raising significant rounds in recent weeks. TechCrunch

The Anthropic appeals court ruling is the one to watch in the coming weeks. If the court sides with Anthropic and overturns the DOD’s supply chain risk designation, it will mark the first time a US court has directly constrained how the federal government can restrict commercial AI models from defence procurement. Separately, watch whether Gemini 3.5 Flash pricing lands competitively enough to pull developer workloads away from OpenAI’s API, and whether Google’s $100 AI Ultra tier sees meaningful take-up once Gmail Live rolls out this summer.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.