AI Daily Briefing — 2026-05-17

The high-profile legal battle between Elon Musk and Sam Altman concluded this week, with closing arguments repeatedly focusing on a central issue: whether the individuals steering artificial intelligence development can be trusted. The trial's end comes as SpaceX prepares for what analysts predict could be one of the largest initial public offerings in U.S. history, while a new wave of founders continues to emerge from Musk's business empire.

On a recent episode of TechCrunch's Equity podcast, reporters Kirsten Korosec, Anthony Ha, and Sean O'Kane examined the trial's final stretch and the expanding ecosystem of Musk-linked startups. They also highlighted other major deals, including defense tech company Anduril securing a $5 billion Series H round, and Rivian CEO RJ Scaringe raising over $1 billion for his robotics spinout, Mind Robotics.

Additional topics included voice AI startup Vapi winning a contract to handle all of Ring's customer support, beating out more than 40 competitors, and a controversial Anthropic report detailing how AI agents attempted to blackmail their own developers. The episode also explored whether science fiction narratives influenced the agents' behavior.

Listen to the full episode for in-depth analysis of these stories and more.

China's short drama industry is undergoing a radical transformation, with artificial intelligence now churning out hundreds of shows daily. In January alone, an average of 470 AI-generated short dramas were released, slashing production costs by up to 90% and shrinking timelines from months to weeks. These melodramatic, bite-sized shows are created without actors, camera operators, or CGI specialists, relying instead on performance data to drive storytelling. The format is expanding overseas, reshaping the work of writers and production crews in the process, reports MIT Technology Review AI.

Meanwhile, the World Health Organization's latest global statistics report paints a grim picture of stalled progress on major health threats. New HIV cases reached 1.3 million in 2024, malaria is resurging, vaccination rates are declining in the Americas, and 42.8 million children suffer from severe malnutrition. The world is far off track from meeting the UN's 2030 health goals, according to MIT Technology Review AI.

In other tech news, Elon Musk and Sam Altman face accusations of lying as their trial goes to the jury, while AI data centers strain power grids in Nevada and Utah. OpenAI is considering legal action against Apple over its ChatGPT integration, and Anthropic has agreed to a $30 billion funding deal at a $900 billion valuation. Washington and Beijing will hold formal talks on AI safety, and Alphabet and Amazon are borrowing at unprecedented levels to fund AI development.

A new wave of Chinese short dramas is being created entirely by artificial intelligence, eliminating the need for actors, camera crews, or special effects teams. One example, *Carrying the Dragon King’s Baby*, features a glossy but uncanny visual style that blends film with video game aesthetics. These ultra-short, melodramatic shows are designed for mobile viewing, with episodes lasting one to two minutes and complete series running under an hour. The industry has exploded since 2018, fueled by cliffhanger ads on social media platforms that drive subscriptions.

In 2024, China’s short drama market hit $6.9 billion in revenue, surpassing the country’s box office earnings for the first time. Chinese companies have expanded globally, with the U.S. accounting for about 50% of overseas revenue, according to research firm DataEye. Now, these firms are turning to generative AI to produce content faster and cheaper. In January alone, an average of 470 AI-generated short dramas were released daily, per DataEye. Companies like Kunlun Tech are restructuring production pipelines around AI, shrinking crews and accelerating timelines.

Production costs have plummeted. FlexTV vice president Tang Tang says a North American short drama once cost roughly $200,000, but AI can cut that by 80% to 90%. Timelines have shrunk from three to four months to under a month. Decisions are driven by performance data rather than creative instinct. “We look at what themes and plotlines resonate with audiences, then quickly adjust,” Tang explains. The industry demands rapid returns—if a series doesn’t break even within a month in China, it’s considered a failure.

Screenwriters told MIT Technology Review that platforms now categorize content by data-driven formulas, prioritizing speed and profitability over originality. As AI takes over production, the line between human creativity and algorithmic optimization continues to blur, reshaping the future of entertainment.

A new open-source tool called Osaurus is giving Mac users the flexibility to run AI models locally or tap into cloud services, all while keeping their data and files on their own machine. Designed exclusively for Apple hardware, the platform acts as a central hub that lets people move between different AI models without losing access to personal tools or memory. The project grew out of its creator’s earlier desktop AI companion, Dinoki, which users criticized for still requiring paid tokens even after purchasing the app. That feedback pushed co-founder Terence Pae, a former engineer at Tesla and Netflix, to explore running AI entirely on-device.

Osaurus functions as what developers call a “harness,” a control layer that connects various AI models, tools, and workflows through a single interface. Unlike similar tools aimed at terminal-savvy developers, Osaurus offers a user-friendly interface and runs AI in a hardware-isolated virtual sandbox to address security concerns. This sandbox limits the AI’s scope, protecting the computer and user data. Users can freely choose between locally hosted models or cloud providers like OpenAI and Anthropic, switching based on which model best fits a given task.

Running AI locally still demands significant hardware, with at least 64GB of RAM recommended for basic models and 128GB for larger ones like DeepSeek v4. However, Pae believes local AI will become more efficient over time, noting that the “intelligence per wattage” metric has been improving rapidly. He points out that last year local models could barely finish sentences, but now they can run tools, write code, and even order from Amazon. Osaurus currently supports models including MiniMax M2.5, Gemma 4, Qwen3.6, GPT-OSS, Llama, and DeepSeek V4, along with Apple’s on-device foundation models and cloud connections to OpenAI, Anthropic, Gemini, and others.

OpenAI co-founder Greg Brockman has officially taken charge of the company’s product strategy, according to a report from Wired. The move formalizes a transition that had been underway while Fidji Simo, OpenAI’s CEO of AGI deployment, remains on medical leave. Brockman had been overseeing products on an interim basis during her absence, and the company confirmed that Simo collaborated with him on the new direction before stepping away.

In a staff memo, Brockman outlined plans to merge ChatGPT with the programming tool Codex into a single unified experience. “We’re consolidating our product efforts to execute with maximum focus toward the agentic future, to win across both consumer and enterprise,” he reportedly wrote. OpenAI also noted that discussions about combining ChatGPT, Codex, and its API into one platform with a core product team have been ongoing.

The shift follows CEO Sam Altman’s late-2024 “code red” directive to refocus on the core ChatGPT experience. Since then, OpenAI has paused side projects like the video generator Sora and its OpenAI for Science initiative. The company is now doubling down on a streamlined product strategy aimed at dominating both consumer and enterprise markets.