AI Daily Briefing — 2026-05-22

Tech researchers are taking legal action against the Trump administration over a visa restriction policy they argue threatens online safety and free speech. The policy, announced by Secretary of State Marco Rubio, targets foreign-born workers involved in content moderation and disinformation research. In a new lawsuit, the researchers claim the rule violates due process and speech rights, potentially reshaping how platforms handle hate speech and propaganda. The case could set a precedent for online safety efforts, as detailed by Eileen Guo.

Meanwhile, climate tech companies are shifting focus to critical minerals amid waning support for decarbonization under the Trump administration. Boston Metal, known for cleaning up steel production, has raised $75 million to produce critical metals, as exclusively reported by MIT Technology Review. This pivot helps firms survive a tough political climate, with Casey Crownhart exploring how reframing missions toward mineral extraction offers a lifeline.

In AI news, researchers are developing 'world models' to overcome LLM limitations, with advances from Google DeepMind and others. An exclusive MIT Technology Review event today examines this progress, part of their '10 Things That Matter in AI Right Now' list. Elsewhere, SpaceX files for a record IPO, Nvidia reports strong revenues despite China losses, and Samsung averts a strike over AI profit-sharing.

A biotech company, Colossal Biosciences, announced it has developed a fully artificial eggshell using 3D-printed plastic cups to grow chicken embryos. The chicks were observed attempting to hatch from these transparent containers at the company's Dallas facility. This innovation is part of Colossal's broader mission to resurrect extinct bird species, such as the dodo and the giant moa. While some scientists caution that the company may be exaggerating the significance of this achievement, the technology could eventually pave the way for artificial womb development.

In other news, Elon Musk lost his high-profile lawsuit against OpenAI, which accused cofounders Sam Altman and Greg Brockman of deceiving him about the organization's nonprofit status. According to MIT Technology Review AI reporter Michelle Kim, who covered the trial, the courtroom proceedings revealed key insights into the ongoing AI race. Subscribers can access a full discussion of the case and its implications.

Additionally, researchers are exploring cryopreservation of brain tissue, with one scientist studying pieces of a friend's frozen brain stored at -146°C in Arizona. The goal of eventual reanimation remains controversial, but the work could advance organ transplantation techniques. Meanwhile, AI researchers are shifting focus toward world models—systems that understand physical environments—as a way to overcome the limitations of large language models.

Hark, an AI lab focused on creating a universal personal assistant, has raised $700 million in a Series A funding round, reaching a $6 billion post-money valuation. The massive investment, led by Parkway Venture Capital and backed by firms like AMD Ventures, ARK Invest, and Salesforce Ventures, underscores the race to build the first must-have consumer AI product. Despite the hefty sum, the company has revealed few details about its technology, which aims to serve as an all-encompassing interface for the digital world.

Founded in late 2025 by Brett Adcock, the entrepreneur behind Figure.AI and Archer Aviation, Hark plans to release its first multimodal AI models this summer. These will power a personal platform compatible with existing services, followed by custom hardware devices. The funding will be used to hire top talent in hardware, design, and AI research, as well as to secure computing power and components. The company currently employs 70 people and operates a data center with Nvidia B200 GPUs.

Hark’s director of design, former Apple executive Abidur Chowdhury, noted that most AI products today focus on helping developers, not everyday users. He believes Hark’s unique focus on native hardware and user-friendly interfaces could fill that gap. However, significant challenges remain, particularly around privacy and context—how to make an AI assistant useful without intruding on users’ personal lives. When asked about this, Chowdhury offered only a cryptic smile, hinting at future innovations.

President Donald Trump has decided to postpone signing an executive order aimed at bolstering AI security, expressing reservations about its potential to stifle progress. The order would have required government agencies to evaluate artificial intelligence models before their public release. Trump told reporters at the White House that he found certain elements of the proposal unsatisfactory, fearing they could hinder the nation's competitive edge. "We're leading China, we're leading everybody, and I don't want to do anything that's going to get in the way of that leading," he stated.

The delay also stems from logistical challenges, as reports indicate that not enough prominent tech CEOs could attend a planned signing ceremony on short notice. The event was seen as a key photo opportunity for the administration. The executive order was designed to address security risks highlighted by recent AI releases, including Anthropic's Mythos and OpenAI's GPT-5.5 Cyber, both capable of rapidly identifying and exploiting vulnerabilities. It would have tasked the Office of the National Cyber Director with creating a pre-release evaluation process.

A major point of contention was a provision requiring AI companies to share advanced models with the government 14 to 90 days before launch. Trump expressed concern that this language "could have been a blocker" for innovation. The decision to delay reflects a broader tension between regulating emerging technology for safety and maintaining U.S. leadership in the global AI race.

Spotify has unveiled a new tool that lets authors create audiobooks using artificial intelligence, powered by voice startup ElevenLabs. Announced during the company’s Investor Day, the feature will launch in beta this June by invitation only, initially supporting English. Authors who use the tool are not locked into an exclusive deal, meaning they can distribute their AI-narrated books on other platforms as well. The move builds on an existing partnership between Spotify and ElevenLabs, which previously allowed writers to submit audiobooks made on ElevenLabs’ platform directly to Spotify.

Spotify is also expanding its “Spotify for Authors” program to support ten additional languages, including French, German, Dutch, and Swedish. The company plans to introduce higher listening limits under its Audiobook+ plans later this year, along with new student and family options, though specific pricing has not been disclosed. To date, Spotify has surpassed one million Audiobook+ subscriptions and expects to generate $100 million in annualized recurring revenue from the program.

In addition to the audiobook creation tool, Spotify is rolling out a natural language search feature for audiobook discovery. This summer, users will be able to create prompt-based playlists for audiobooks, similar to existing options for music and podcasts. The company has aggressively expanded its audiobook catalog to 700,000 titles, boosted listening hours by 60% year-over-year, and noted that more than half of its audiobook listeners started within the last year.