AI News July 15, 2026: Apple Sues OpenAI Over Trade Secrets, Google-Microsoft Forge Enterprise AI Alliance, AI Token Price War Escalates, China's MiniMax Plans 2.7T Open-Source Model, OpenAI Launches GPT-Live Full-Duplex Voice

⚖️ Top 5 AI Stories — July 15, 2026

The AI industry entered a new phase of open conflict this week. Apple fired a legal salvo at OpenAI, accusing the company of systematically poaching talent and stealing trade secrets. On the enterprise front, Google and Microsoft assembled a coalition of software giants to build a shared AI agent protocol aimed squarely at breaking OpenAI and Anthropic’s grip on the market. Meanwhile, AI token prices collapsed as three frontier labs slashed rates in a coordinated land grab for developers. China’s MiniMax announced plans for the largest open-source model ever attempted. And OpenAI rolled out GPT-Live, a voice model that can hold a genuine two-way conversation. Here are the five stories that mattered most.

1. Apple Sues OpenAI for Trade Secret Theft, Alleging 400+ Employee Poaching

On July 11, Apple filed a lawsuit against OpenAI alleging systematic theft of trade secrets, sending shockwaves through Silicon Valley. The lawsuit centers on a striking number: more than 400 former Apple employees now work at OpenAI, many recruited from the teams that design Apple’s custom silicon and on-device AI systems. Apple claims that these employees brought proprietary knowledge about chip architecture, neural engine design, and on-device model optimization that OpenAI used to accelerate its own hardware and consumer device ambitions.

The fallout extended well beyond the courtroom. Elon Musk and Sam Altman spent the weekend trading public accusations on X (formerly Twitter), turning a legal filing into the loudest personal feud in AI. Musk, who has his own acrimonious history with OpenAI, used the lawsuit to renew his attacks on the company’s governance and Altman’s leadership. Altman fired back, defending OpenAI’s hiring practices and questioning Musk’s motives.

The legal battle highlights a deeper tension: as AI companies race to build consumer hardware — OpenAI’s partnership with former Apple designer Jony Ive, SpaceX’s AI device prototypes, and Meta’s AI glasses — the talent war between Big Tech and AI labs has escalated from aggressive recruiting into open litigation. Apple’s lawsuit could set a precedent for how trade secret law applies to the AI industry’s talent drain, and the outcome may reshape hiring practices across the entire sector.

Source: Build Fast With AI · Unrot.co

2. Google and Microsoft Forge Enterprise AI Alliance Against OpenAI and Anthropic

In a move that could reshape the enterprise AI landscape, Google, Microsoft, Salesforce, Snowflake, and ServiceNow have agreed to support a shared technical standard for connecting AI agents to business software, according to The Information. The alliance is framed explicitly as a counterweight to the growing dominance of OpenAI and Anthropic in the enterprise market.

The coalition aims to build an open protocol — sometimes described as an “AI agent API layer” — that would allow AI agents from any provider to interact with enterprise systems in a standardized way. This would reduce lock-in to OpenAI’s or Anthropic’s proprietary ecosystems and give customers more flexibility to switch between models. For Google and Microsoft, which both offer cloud AI infrastructure, the strategy is clear: commoditize the model layer and compete on platform, tooling, and enterprise relationships instead.

The timing is significant. OpenAI’s GPT-5.6 and Anthropic’s Claude Fable 5 have dominated enterprise adoption metrics in 2026, with Fortune 500 companies increasingly building critical workflows around these models. By creating a shared standard, the alliance hopes to prevent any single AI lab from becoming the indispensable layer of enterprise software. Whether CIOs will embrace yet another standard remains to be seen, but the competitive pressure on OpenAI and Anthropic is undeniable.

Source: Build Fast With AI · The Information

3. AI Token Price War: GPT-5.6, Grok 4.5, and Meta’s Muse Spark Trigger Race to the Bottom

The AI inference market has entered a full-blown price war. OpenAI’s GPT-5.6 Sol launched at $5 per million input tokens and $30 per million output tokens — directly undercutting Anthropic’s Claude Fable 5 at $10/$50 by 50%. SpaceXAI’s Grok 4.5, released July 8, came in even more aggressively, priced at $5/$25, positioning itself as a “faster, cheaper Claude rival.” And Meta’s Muse Spark 1.1, also launched July 9, dramatically undercut both at just $1.25 per million input tokens and $4.25 per million output tokens — roughly a tenth of Anthropic’s Fable 5 pricing.

For developers and enterprises, this is a bonanza. API costs that dominated AI budgets just months ago are collapsing. But for AI labs, it signals that the era of premium model pricing is ending. OpenAI’s GPT-5.6 Sol claims to deliver state-of-the-art coding performance — scoring 80 on the Coding Agent Index, 2.8 points above Fable 5 — while using less than half the output tokens and costing about one-third less. The message: better models at lower prices.

Meta’s aggressive pricing strategy is particularly notable. By pricing Muse Spark 1.1 so far below competitors — and offering $20 in free credits per account — Meta is effectively subsidizing developer adoption to build ecosystem loyalty around its new developer API portal. SpaceXAI’s Grok 4.5, meanwhile, leverages compute capacity leased from the same SpaceX infrastructure that serves competitors like Anthropic and Google, creating a remarkable dynamic where rivals are literally paying each other.

Source: Euronews · Constellation Research · Medium — nlplanet

4. China’s MiniMax Plans 2.7-Trillion-Parameter Open-Source Model

Chinese AI startup MiniMax is developing a large language model with 2.7 trillion parameters, Reuters reported on July 8 — what would be the largest open-weight AI model released by any company, Chinese or otherwise. The model, reportedly called M3 Pro, is six times the size of MiniMax’s current flagship and would dwarf every other open-source model on the market.

MiniMax plans to open-source the model as early as Q3 2026, according to The Information, citing two people familiar with the plans. This comes at a pivotal moment: Chinese open-source models from MiniMax, Zhipu AI (makers of GLM-5.2), and DeepSeek are gaining significant traction in the U.S. and global markets as lower-cost alternatives to proprietary American systems. GLM-5.2, released under an MIT license with a 1-million-token context window, has already climbed to #2 on Code Arena — the highest-ranked openly available model.

The open-source dynamic has been amplified by U.S. government actions. The Trump administration ordered Anthropic’s most advanced models (Fable 5 and Mythos 5) blocked for foreign nationals, and on the same day, Zhipu announced GLM-5.2 with no usage restrictions. Chinese labs are positioning open-source as a strategic counter to U.S. export controls — if American frontier models are restricted, Chinese open models fill the gap for free. For developers worldwide, this means access to increasingly capable models at a fraction of the cost of proprietary alternatives.

Source: Reuters · The Next Web · The Decoder

5. OpenAI Launches GPT-Live: Full-Duplex Voice AI That Talks Like a Human

OpenAI launched GPT-Live on July 8, a new generation of voice models designed to make talking with AI feel like a genuine conversation. Built on a full-duplex architecture, GPT-Live can listen and speak simultaneously — meaning users can interrupt it mid-sentence, pause to think, or engage in rapid back-and-forth without awkward delays. During conversations, it shows attentiveness with natural acknowledgment phrases like “mhmm” or “yeah.”

What sets GPT-Live apart is its ability to multitask within a conversation. For questions requiring web search, deep reasoning, or complex work, GPT-Live delegates to OpenAI’s latest frontier model (currently GPT-5.5) behind the scenes, then brings the results back into the conversation when ready. While working on a task, it can continue talking with the user, maintaining conversational flow rather than falling silent.

The launch also included three new API models: GPT-Realtime-2 (with GPT-5-class reasoning), GPT-Realtime-Translate (live translation across 70+ input languages into 13 output languages), and GPT-Realtime-Whisper (streaming speech-to-text). These models give developers the building blocks for a new class of voice applications — from real-time meeting translation to AI receptionists that handle complex multi-turn conversations. OpenAI emphasized that GPT-Live is designed for utility, not companionship, a notable distinction as AI voice products face growing scrutiny over emotional attachment and dependency.

Source: OpenAI · TechCrunch · MacRumors

🔑 Key Takeaways

The AI industry is simultaneously colliding and fragmenting. Apple’s lawsuit against OpenAI signals that the talent war has crossed into legal warfare. The Google-Microsoft enterprise alliance shows incumbents fighting back against AI-native companies. The token price war is making frontier AI dramatically cheaper for everyone. China’s open-source surge is reshaping global access to powerful models. And voice AI has crossed a threshold from novelty to genuine utility. The common thread: AI competition has moved from model benchmarks to courts, alliances, pricing strategies, and geopolitical chess.

This article was automatically compiled from multiple sources on July 15, 2026. All facts are attributed to the cited sources. For the latest developments, follow the source links above.