Pay packages in the AI industry are some of the most generous in tech. Salaries in the U.K. government… not so much.

So Jade Leung took a big pay cut when, in October 2023, she quit her job at OpenAI to become chief technology officer to the U.K. government’s new AI Safety Institute (AISI). Established in late 2023, the U.K. AISI has in a short time become the leading government body, anywhere in the world, dedicated to testing the safety of the most powerful AI systems.

The move was worth it, Leung says, because of the opportunity to shape the way that AI systems are governed. “You really want a public interest body that is genuinely representing people to be making those decisions,” she says. “By definition, they need to be [taken] outside of the organizations developing” AI models.

Leung’s role at the U.K. AISI has largely been focused on building it into a body that can carry out state-of-the-art safety research. Alongside staffing up the nascent organization, her key responsibility is designing and overseeing evaluations that can test whether AI models pose the risk of being used to facilitate cyber, biological, or chemical attacks. The organization has also been developing a more speculative set of tests designed to check whether AI systems have the ability to escape the control of their designers. The U.K. government has negotiated early access to new models from leading AI companies to carry out safety testing on them; the AISI has completed two such tests already, one of them on Anthropic’s Claude 3.5 Sonnet model. “We are currently, I would say, the best single place that you could go to to have your models evaluated across a range of frontier AI risk areas in a fairly robust and in depth way,” Leung says.

Buy your copy of the TIME100 AI issue here

It’s hard to imagine what modern life would look like without Google. Its search business prints hundreds of billions of dollars in yearly revenue, and starting over two decades ago, Google began channeling some of that money toward AI research. Its industry-leading scientists were responsible for many of the breakthroughs that drove the field to its current inflection point. And yet the product that in late 2022 kick-started today’s AI boom, ChatGPT, came from a startup backed by Google’s major competitor, Microsoft. Suddenly Google was no longer the symbolic leader of the AI race, but instead playing catch-up.

Google’s CEO Sundar Pichai, who joined the company in 2004 and was appointed to the top job in 2015, took that hurdle in stride. Google wasn’t the first to build a search engine, he points out, but was the first to build one good enough to attract the lion’s share of the market. The same for browsers. Email. Maps. His point: it matters less whether Google is first, and more that its version is the best. The U.S. Department of Justice takes an alternative view: that Google’s search is a monopoly upheld by illegal anti-competitive actions. On Aug. 5, a judge ruled in favor of that argument; Pichai says Google plans to appeal.

Last year was a huge one for Sam Altman. The OpenAI CEO rocketed to household name status, was sought by dozens of world leaders for his audience, and saw his personal wealth balloon, all thanks to the success of ChatGPT. But just before Thanksgiving, the rocket engine faltered. OpenAI’s nonprofit board fired Altman, saying he had been dishonest with them. Altman eventually saw off the board’s threat, and was back in the top job by the time he sat down for home-cooked vegetarian pasta on Thanksgiving. But to many observers—if not for his many allies in Silicon Valley—the events had tarnished Altman’s golden year.

In 2024, as well as shepherding OpenAI through several new products including a voice assistant, a synthetic video generator, and an AI search engine, Altman resumed his project of smoothing the ground for OpenAI’s future plans. In February, it emerged he was seeking up to $7 trillion from investors in an audacious attempt to build a new manufacturer of the chips necessary to build cutting-edge AI, seeking to loosen some of Nvidia’s stranglehold of the market. And in July, he published a manifesto of sorts in the Washington Post, arguing that for democracies to succeed against authoritarianism, they must improve their cybersecurity, invest more deeply in AI hardware, and collaborate with China on setting global norms for AI deployment. 

As the Director of the Cyberspace Administration of China (CAC), the country’s leading AI regulator, Zhuang Rongwen faces a seemingly impossible task: imposing the nation’s censorship practices on unpredictable generative AI technology without stifling domestic innovation.

That did not stop Mr. Zhuang from making history in August 2023, when less than a year after ChatGPT was released, China became one of the first countries to implement binding regulation targeting generative artificial intelligence. The law requires, among other things, for developers to obtain government approval before deploying models publicly. Since then, the government has approved over 40 Chinese language models, including those of newly-minted Chinese AI unicorns like Zhipu and Baichuan. China is second only to the United States in the number of large-scale AI models originating from the country, accounting for 25% of such models. 

A provision in the law states that AI developers must “Uphold the Core Socialist Values.” Perhaps leading by example, a research arm of the CAC created its own LLM in May. Dubbed ‘Chat Xi PT,’ the model was trained on the political doctrine of President Xi Jinping as well as other official literature. For now, the model is for internal use only.

But the CAC under Zhuang has also been instrumental in curtailing the power of China’s biggest tech companies. In the summer of 2022, the regulator issued a $1.2 billion fine to China’s leading ride-hailing company, Didi, for violating data security laws. The massive fine was a part of a multi-agency crackdown on China’s tech industry that wiped out more than $1 trillion in value from the sector. 

While the crackdown has since eased, there is no guarantee it will last. Zhuang’s decisions will help shape whether China can keep pace with its western counterparts and realize its aspirations of becoming an AI powerhouse. 

Dr. C.C. Wei is in charge of the world’s largest contract semiconductor manufacturer, Taiwan Semiconductor Manufacturing Company (TSMC), having served its CEO since 2018, and its Chairman since June 2024.

TSMC, whose clients include Apple, NVIDIA, and AMD, has taken on new importance amidst the ongoing AI boom—it is the only company in the world currently capable of fabricating the cutting-edge chips that power the most advanced AI systems. Its market value has nearly tripled over the past six years.

The company’s unparalleled manufacturing capabilities, combined with the strategic importance of its product, puts it in a fraught position in relation to the tensions between the U.S. and China. Wei is careful about what he says about this relationship in public. Reflecting on his career during a talk at Yale, his alma mater, Wei said with an apologetic smile “I was not allowed to give a speech without going through my legal department…I don’t make a public comment on China,” going on to note that many of TSMC’s customers were from China.

Wei has been at the company since 1998. Under his leadership, TSMC has initiated a significant global expansion beyond Taiwan, with plans for new advanced chip factories in the United States and Germany, and ongoing construction in Japan.

TSMC has been one of the beneficiaries of billions of dollars of subsidies provided under America’s CHIPS Act. To date, they have been awarded over $6 billion to support the construction of a factory in Arizona. There’s a “huge amount of talented people” in America, Wei told Yale students. He believes the company’s westward expansion is an opportunity for it to recruit and build towards the future.

Buy your copy of the TIME100 AI issue here

Americans already know ByteDance as owner of social media phenomenon TikTok, which Congress ruled in March would be banned in the U.S. unless its Beijing-based parent divests. However, ByteDance is also making huge strides in AI under the leadership of Liang Rubo, who replaced his former roommate and co-founder, Zhang Yiming, as chief executive in 2021. In May, the company released an AI model, dubbed Doubao, which has overtaken Baidu’s ERNIE as China’s most popular chatbot and costs 99.8% less to run than OpenAI’s ChatGPT4, according to ByteDance. But if the company has been given a rough ride by U.S. lawmakers over TikTok, don’t expect any softening when it comes to broader AI applications, with Washington determined to maintain any competitive advantage via export controls. In response, ByteDance is investing $2 billion to develop an AI hub in Malaysia and is negotiating with Broadcom to design an advanced AI chip. Seizing on the potential of AI has become something of an obsession for Liang, who went viral for lambasting colleagues for being slow to identify the technology’s potential. “This is an era of great change,” Liang told a company meeting in January, urging staff to “escape the gravity of mediocrity.”

Wang Xiaochuan had been marked for greatness from a very young age. The Chengdu native won first prize in China’s National High School Maths Competition aged 14 and a gold medal in the International Olympiad in Informatics at 17. He founded his first company while still a student at China’s prestigious Tsinghua University and became a vice-president of the nation’s second most popular search engine, Sogou, at just 27. It was after stepping down as CEO of Sogou in 2021 that Wang focused his undoubted intelligence on the artificial variety. In April 2023, he founded Baichuan AI, and within six months had already raised $300 million from investors like Alibaba and Tencent, making it one of the fastest companies to achieve unicorn status ever. Wang hasn’t disappointed; in May, Baichuan released its fourth generation large language model, which the company says outperforms competitors in several key metrics. The Beijing-based company—valued at $2.7 billion today—has also unveiled an AI assistant, Baixiaoying, which has the unique ability to ask users follow-up queries to figure out their intentions and better tailor responses. “We're the only one doing this—to guide and inspire a user to clearly articulate what they need,” Wang told an industry event in May.

Lisa Su is acutely aware that technology is all about making the right bets. In her 10 years at the helm of chipmaker Advanced Micro Devices (AMD), she has steered one of Silicon Valley’s greatest turnarounds by focusing on its strengths and strategic dealmaking. When she took over at the company in 2014, its share price was languishing around $3. By leaning into making central processing units (CPUs) for laptops and PCs, and graphics processors, used in gaming consoles and PCs, Su brought AMD onto more stable footing, strengthening tactical partnerships with companies including Sony and Microsoft. In recent years, AMD has closed the $49 billion purchase of competitor Xilinx—the largest semiconductor deal ever—as well as the $1.9 billion acquisition of data center networking company Pensando.

Deepfakes as a service: that’s Synthesia’s business model. But it’s not what it sounds like. The London-based AI company offers businesses a platform to turn any document or webpage into an engaging video hosted by a realistic AI avatar (resembling either you the user, or an actor whose likeness Synthesia has paid to license) that can change its expressions, body language, and intonations based on the text it’s reading. In June, the company launched a service allowing customers to create an avatar of themselves using a laptop webcam in as little as 5 minutes. Victor Riparbelli, Synthesia’s CEO, explains that most humans learn more easily from watching a video than they do from reading. AI-generated video is glitchy and not ready for prime-time, he says, but Synthesia is targeting use cases where viewers don’t tend to care. “Our product is not made for making the next Super Bowl ad,” Riparbelli says. “It’s much more made for [clients who say], ‘We have a help-center article, make a quick video that explains the content.’”

Synthesia is also battling against the seedier ways that people are using AI-generated video. Political deepfakes, showing election candidates saying things they never said, have surfaced in elections in India and Slovakia so far this year, although with fewer adverse effects than some had anticipated. More devastatingly, the technology has been used to produce an epidemic of sexualized harassment of women and girls, their likenesses superimposed onto explicit videos without their consent. Synthesia lobbied the U.K. government to ban the creation and sharing of sexually explicit deepfakes–a law that came into effect in April. On its own platform, Synthesia requires users to upload a verification video of themselves reading a script, to ensure that individuals can’t make AI avatars of other people. “I would hope most people agree that non-consensual deepfakes, nobody wants that,” says Riparbelli. “And I think it’s very positive that we’re making policies against that.”

Buy your copy of the TIME100 AI issue here

Online, Aravind Srinivas displays swagger. In person, the 30-year-old has the demeanor of a man not used to being the center of attention. Sitting down for an interview in May, the CEO of the AI “answer engine” Perplexity is told that his potential inclusion in the TIME100 AI was not yet final and the decision could still change in the months ahead. “OK,” he says. “So as long as I don’t f-ck up…”

By then, Perplexity had already drawn the ire of online publishers for its business model: using AI to answer users’ questions by summarizing websites, diverting ad revenue from those sites. In June, the company launched a new feature called “Perplexity Pages,” that creates AI-generated reports in response to users’ queries. The feature appeared to have plagiarized sections of reporting from multiple publications, while barely or inaccurately citing its sources.

One of the most oft-dreamed applications of artificial intelligence is using it to pioneer new screenings, treatments, and interventions for diseases. This is precisely what Daphne Koller is trying to do at Insitro, where she’s founder and CEO.

Insitro uses machine learning tools to scan genetic samples from people with diseases like ALS, cancer, and tuberous sclerosis. It then tries to identify causal mechanisms that human researchers may have missed. Koller describes it as a collaborative environment, where people with backgrounds in fields like machine learning, metabolic disease research, and statistical genetics often work together.

From the outside, Koller’s path may not look like a straight one: she was a faculty member of Stanford’s computer science department and co-founded the online learning company Coursera in 2012, before founding Insitro in 2018. 

“I've had this increasing sense of urgency to make an actual direct impact in the world,” Koller says. While observing the beginning of exponential growth in AI capabilities in 2016, she felt an urge to help unlock its real-world potential.

“You could use [AI] for things like making a better sales chat bot—and I'm not trying to dismiss that, but I think you can also come up with more aspirational use cases for this incredible technology,” she says. 

Insitro’s research into ​​nonalcoholic fatty liver disease, Koller says, has advanced to animal trials. The company also plans to apply for a clinical trial for a human drug in the coming months. Six years in, Koller is excited by Insitro’s progress, but says it’s crucial not to play into the AI-hype cycle that’s so common in the tech world.

“What we're doing is really hard—intervening in human biology in a way that is both safe and efficacious,” Koller says. “So it's important to not make extravagant promises that are just not the right ones for this space.”

Buy your copy of the TIME100 AI issue here

With each new variant of SARS-COV-2—not to mention annual variants of the flu—another wave of people are sickened, and their lives disrupted. Viruses mutate and evolve, keeping ahead of vaccines. EVEScape aims to predict those evolutions and mutations by using artificial intelligence. The project, which is co-led by Sarah Gurev, a doctoral student at Harvard Medical School in the Debora Marks Lab, uses AI to process data on viral evolution over an extended period, alongside information on the biology and structures of different viruses. The team, including co-leaders Nicole N. Thadani, Debora Marks, Pascal Notin, and Noor Youssef, published a study last fall showing that if the tool had been available at the start of the pandemic in 2020, it would have been able to predict the most concerning variants before they appeared. Gurev designed the AI models that underpinned those predictions.

Gurev and her colleagues also collaborate with labs that can evaluate vaccines and therapies that can stop future variants of a virus from wreaking havoc. “The direct partnership between computationalists and experimentalists will really drive AI to become impactful in different biological applications,” she says.

Gurev and her collaborators are now using the tool to keep up with the Lassa and Nipah viruses, which lead to serious illnesses in West Africa and South Asia respectively, and are more virulent that SARS-COV-2 but much less studied. Vaccine design is typically slow and laborious, but AI may speed up and streamline the process. Gurev says the tool can also help researchers design broader coronavirus vaccines that would be “future proof.”

This March, Mustafa Suleyman left his startup Inflection AI to become the CEO of Microsoft AI, a newly-formed unit focused on the company’s consumer artificial intelligence research and products, like Copilot, Bing and Edge. Almost all of Inflection’s staff followed Suleyman to Microsoft from the startup that was valued at $4 billion. At the same time, Microsoft inked a $650 million deal with the startup to access its AI models.

Suleyman took a humanistic approach to technology with his strategy at Inflection. The company sought to create the first “emotionally intelligent” chatbot. In a TED Talk, a month after joining Microsoft, he described himself as someone “who’s always cared deeply about [technology’s] ethics.” Back in 2018, while at DeepMind—the Google-owned AI lab he co-founded—Suleyman signed an open letter endorsing a ban on lethal autonomous weapons.

Yet he has become embroiled in some ethical tangles of his own. He departed DeepMind in 2019 following accusations that he had bullied his staffers. He later apologized for his behavior. More recently, Microsoft’s arrangement with Suleyman and Inflection came under regulatory scrutiny, with the U.K. Competition and Markets Authority launching an investigation into whether it constituted a merger. On Sept. 4, the watchdog cleared the deal, saying that, while it qualified as a merger, it did not cause competition concerns. Microsoft and Inflection had said that they would cooperate with the watchdog. Microsoft added that it believes hiring talent promotes competition and shouldn’t be treated as a merger.

Buy your copy of the TIME100 AI issue here

It’s been a tumultuous 12 months for Mira Murati, one of the driving forces behind OpenAI’s meteoric rise. As the company’s chief technology officer, Murati worked behind the scenes to guide models like ChatGPT and DALL-E to mainstream adoption. But last November she was thrust into the spotlight when she was appointed interim CEO following the controversial—and brief—ouster of Sam Altman. According to a New York Times report, Murati had shared her concerns about Altman’s leadership with the board. But after a widespread pushback from employees and investors, Altman was reinstated to the top post, with Murati’s full support. (Murati wrote in a statement in March that her comments to the board about Altman were “all feedback Sam already knew,” adding “I fought their actions aggressively and we all worked together to bring Sam back.”)

In the months since, Murati has played a larger public role at the company. In May, she introduced the company’s new flagship model, GPT-4o, which impressed users with its sophisticated real-time voice interactions. She showcased the company’s progress on SearchGPT, a search engine, and Sora, a text-to-video model. At the Metropolitan Museum of Art’s annual Met Gala, she spoke about OpenAI’s chatbot installation at the museum.

But Murati has also had some high-profile stumbles: When she was asked whether the training data for OpenAI’s video generation model Sora came from copyrighted sources like YouTube, she answered evasively. She provoked another firestorm in June when she claimed that AI would replace some creative jobs that she said perhaps “shouldn’t have been there in the first place.” 

These comments have contributed to a rocky year in the public perception of OpenAI. Meanwhile, Murati and her team are forging ahead with their next generation text model, which Murati predicts will possess the intelligence of a PhD student for some specific tasks—and could arrive in a year.

Buy your copy of the TIME100 AI issue here

*Disclosure: OpenAI and TIME have a licensing and technology agreement that allows OpenAI to access TIME's archives.

Without ASML, the world’s only producer of manufacturing equipment for cutting-edge semiconductors, the AI boom can't happen. That’s how Dutch giant ASML’s former CEO put it to Bloomberg News in January. Normally, one can dismiss such comments from a chief executive as hyperbolic marketing speak. But in ASML’s case, they might be right.

This spring, the company elevated Christophe Fouquet, its former chief business officer to the chief executive position. Fouquet is now responsible for overseeing how ASML responds to the growing demand for semiconductors. It is the only company in the world that builds the machines needed to produce the AI chips designed by heavyweights such as Nvidia.

Its machines, which cost up to $380 million each, are used by semiconductor manufacturers like Taiwan’s TSMC to etch unfathomably small patterns—measured in single-digit nano-meters—in pieces of silicon. Achieving this level of precision involves hitting tiny droplets of molten tin with lasers to emit extreme ultraviolet light, in a technique known as EUV lithography. 

Fouquet also has to navigate the company as it finds itself caught amidst growing U.S.-China tensions. Washington-led export restrictions imposed in January, aimed at curbing China’s access to advanced semiconductors, are expected to reduce the company’s sales in the country by 15%, and the Biden administration is considering tighter controls. Fouquet has previously stated he believes decoupling the global semiconductor supply chain would be “incredibly expensive” if not impossible. He’s also warned that cutting off China may only push the country to develop its own ASML-alternative.

Beginning last year, a wave of tech giants like Meta, X, Microsoft, and Zoom quietly have updated their terms of service to allow themselves to use customer data to train AI. Since 2014, Proton, the Swiss privacy company, has battled threats to their users’ data from authoritarian regimes and hackers. Now, the threat was coming from within the tech industry itself. For 32-year-old Anant Vijay Singh, Proton’s Product Lead, the mission became urgent. “Challenging Big Tech is hard,” he said. “But that's what excited me.”

Singh is pioneering an approach that harnesses AI to enhance privacy, not compromise it. Under his leadership, Proton introduced three privacy-preserving tools. Sentinel, launched last August, combines AI with human expertise for advanced account protection. In July 2024, Docs, a collaborative editor challenging Google Docs, and Scribe, an AI email writing assistant, followed. Both keep users’ data fully encrypted, enabling workflows free from data-hungry giants like Google, Microsoft, and OpenAI. “We don’t just throw AI on everything,” said Singh. “We saw a use-case that we could solve.”

In a world where the wisdom “if you’re not paying for it, you are the product” rings more true every year, Singh's work at Proton offers a compelling alternative—and it's gaining traction. With more than 100 million users worldwide, Proton's growth suggests a hunger for privacy-focused tech.

With just a few minutes of an audio sample, ElevenLabs can clone your voice.

The unicorn start-up, co-founded by two Polish friends, Piotr Dabkowski and Mateusz Staniszewski, has developed a model that can generate a variety of realistic voices from text or pre-existing speech, and be used for dubbing, so that one can appear to sound fluent across 29 languages. 

While only two years old, the company is valued at $1.1 billion and received $80 million from Andreessen Horowitz and others in January. The startup has since partnered with several large brands including HarperCollins, the Washington Post, and TIME. 

ElevenLabs’s model is more robust, has a lower latency, and higher quality than competitors. Dabkowski expects the model to soon hit a “quality ceiling,” where the AI is indistinguishable from any human it mimics.

This poses an obvious safety challenge as it could be used to fraudulent ends, but Dabkowski says that their systems monitor usage to minimize the threat. “We can trace back every single generation, and you need to use your credit card to have access to voice cloning—so it's like using a gun that is connected to the internet,” he says. 

ElevenLabs has developed a speech classifier that can identify whether any audio was generated using their model. The company is also a member of the U.S. AI Safety Institute’s Consortium. Open-source AI audio solutions are a larger concern, Dabkowski adds. Still, some have found ways around some of ElevenLab’s safeguards, and the terrain is constantly evolving, so preventing its misuse may not be so simple.

Buy your copy of the TIME100 AI issue here

*Disclosure: TIME and ElevenLabs have a technology agreement to create audio accessible content.

For the past 10 years, filmmaker Lawrence Lek has been creating science-fiction landscapes that feature neon temples, drones touring abandoned luxury hotels, and empty urban highways.

The throughline is, there are no visible people. His main characters—and crucially, not the villains—are artificial intelligence entities. 

His 2021 film Black Cloud stars a depressed self-driving car that performed its job too well, and finds itself bored and aimless. Meanwhile, his 2017 film Geomancer tells the story of an unfulfilled satellite that returns to earth, longing to become an artist.

These characters deliver poetic, at times morose monologue and dialogue against dreamlike, spacey electronic music.

AI characters, Lek said, can be avatars for outsiders, "or somebody who isn't fully understood yet." He says that treating AI as a threat, in effect, casts it, in human terms, as a fully-fledged adult.  “I actually think it's this adolescent—it's this grumpy teenager who doesn't quite know what they're doing,” Lek says.

As an artist, Lek is an outsider to the AI community, but that perspective allows him to be introspective about the advent of this new technology. Earlier this year, he was awarded the Frieze London 2024 Artist Award. The judges lauded his “essential interrogations into the use of AI and its relationship with the human experience.”

Lek's interest in this storytelling approach was inspired, in part, by growing up in Hong Kong and Singapore, both places that grew and changed rapidly during his childhood.

“It basically didn't seem that science fiction was so much just something that was witnessed in artworks, and in films, and in video games, but it was also something that was part of my everyday life,” Lek says.

A growing number of companies are using chatbots. Einstein GPT–a generative AI platform which rolled out last year, can go beyond questions and answers and help customers with more complex tasks like product returns or refunds. It is a first-of-its-kind Gen AI geared towards customer relations management (CRM), supercharging the work of sales teams and customer service people around the world.

Silvio Savarese, who leads Einstein GPT, says Salesforce has been betting big on generative AI since before ChatGPT's popularity exploded, developing a large language model to help write code and deploying it for internal and external developers. But now, they are moving into agentic AI, which can work autonomously to help automate things like customer service. It can also flag appropriate sales leads and generate marketing emails and copy.

The company has just released an AI benchmark for customer relationship management, which aims to help businesses make more informed decisions when choosing the best large language models for their business. The benchmark can help businesses choose a leaner, more targeted model rather than a huge one, one that can also come with a smaller carbon footprint than larger more power-hungry models, Savarese says.

*Disclosure: TIME co-chair and owner Marc Benioff is CEO of Salesforce.

Federal Trade Commission chair Lina Khan has been a thorn in the side of Big Tech companies since her 2021 appointment, arguing that major players have become too monopolistic and powerful. Over the last year, she’s turned her focus to how these tech giants might be wielding AI to further consolidate their power. In January, the FTC opened an inquiry into the partnerships between Silicon Valley behemoths (Google, Amazon, Microsoft) and rising AI labs (OpenAI, Anthropic), probing whether they might be undermining fair competition. She warned in February that this new technology is in danger of being “co-opted by some of the existing dominant firms to double down on their dominance.”

Khan has sent out many other warning shots to AI companies engaging in questionable practices. She’s declared that any models trained nonconsensually on data from news websites or artists could be in violation of antitrust laws. She has voiced concern about people’s data being used without their knowledge or consent. And she has also proposed new protections against deepfakes. All of these actions, combined with her position of power, make Khan one of the most prominent forces working to combat AI harms today.

Buy your copy of the TIME100 AI issue here

Dan Neely was early to spot the potential problems AI-generated deep-fakes would cause celebrities. The serial entrepreneur says he first noticed crypto mining rigs shifting to generate deep fakes around 2018. Not long after, AI-generated videos of rappers like Jay-Z started showing up singing songs like Billy Joel’s “We Didn’t Start the Fire.”

In 2021, Neely founded Vermillio to help celebrities protect their likeness from being misused. The company bills itself as “the first generative AI platform built specifically to protect the work of content creators.” In January, it signed a deal to protect artists represented by the agency WME, which includes artists like Olivia Rodrigo and actors like Hugh Jackman. 

To protect a client, Vermillio ingests all of their content and builds a “holistic likeness model of who that person is,” Neely explains. The company’s software then scans major platforms for possible matches. Neely hopes that in the age of generative AI, Vermillio will eventually become “the blue checkmark of the internet.” 

Customers have the choice of allowing some types of content, like fan art, while prohibiting others, like deepfake pornography. Vermillio also helps artists charge a license for promotional campaigns that may use their likeness. 

At about $4,000 a month, the company is consciously starting “with the world's most beloved IP and talent,” but it will eventually open-source the underlying technology, because Neely believes protecting your likeness is not just something the most privileged should have access to. “It’s a human right,” he says.  

This summer, as Kendrick Lamar and Drake fired vicious disses back and forth, an unlikely third party ended up creating one of the defining songs of the high-profile rap feud. King Willonius, a little-known New York-based comedian, used AI tools to create “BBL Drizzy,” a pristine soul song that seemed to be unearthed from the 1970s and whose lyrics lightly mocked Drake. Thousands of people then recorded their own remixes, rap verses, guitar solos, or dances over Willonius’s track, sharing them on platforms like TikTok.

For King Willonius—whose real name is Will Hatcher—the success of “BBL Drizzy” served as the gratifying payoff for months of experimentation with AI tools. In 2023, Hatcher was trying to make it as a comedy writer in Los Angeles when the writers strike began, snuffing out the pitch meetings he had lined up with agents and managers. To fill his time and to challenge himself, Hatcher began playing with AI tools for more than eight hours a day: creating movie trailers with Runway, fanciful images with MidJourney, pitch decks with ChatGPT and new songs with Udio and Suno. “Everything unlocked for me: It felt like I now had the resources to create anything I could imagine,” Hatcher tells TIME. 

With his defense technology startup Anduril Industries, Palmer Luckey is betting that the future of military technology lies in advances in software engineering and computing. And with a valuation of $14 billion, investors in the company seem to agree.

Anduril—named after a sword from Lord of the Rings meaning “flame of the west”—has gained a reputation as a disruptor in the defense sector. The company counts the U.S. Department of Defense, the U.K. Ministry of Defence, and the Australian Defence Force among its clients. It has sold more than a dozen autonomous defense and weapons systems to the U.S. since it was founded in 2017.

These systems—which include autonomous drones, rockets, and submarines—are primarily powered by Anduril’s Lattice OS, which it describes as an “AI-powered open operating system” that integrates AI-driven decision-making with flexible, scalable hardware.

Describing Anduril’s ALTIUS drone in an interview with NPR, Luckey explained that the drone has “a Lattice brain” that is able to seek and identify targets and fly into them, “even if they’re jamming you."

Anduril’s tech is quickly spreading across the world. One of its first contracts was to supply AI-driven surveillance towers along the U.S.-Mexico border for the Trump administration. The company’s drones were in Ukraine by the second week of the war, its software continuously updated remotely. Anduril’s newest venture, “Arsenal” aims to use a software platform—and associated manufacturing facilities—to “hyperscale” hardware manufacturing.

Luckey previously spun up one of the first virtual reality startups, Oculus VR, for which he graced a 2015 cover of TIME, before selling it to Facebook for $2 billion in 2014 at the age of 21. He has been described as having a worldview of hyper-techno-optimism and is a Republican donor, but he has said he believes his business is set to benefit regardless of political administration.

When someone is having a stroke, they lose 2 million brain cells every minute. It’s a leading cause of disability and death globally. Chris Mansi, a neurosurgeon turned co-founder of Viz.AI, wants to help people get the care they need much faster.

The company’s algorithms, which analyze a patient’s various tests and scans to help diagnose stroke and other emergencies, can already shave off 66 minutes of the time it takes for stroke patients to receive specialized care. “That is equivalent to reducing disability by one year,” Mansi says. The algorithm’s success has led to Viz.ai deploying it in over 1,600 hospitals.

The company’s platform also facilitates communication with specialists at different hospitals, helping ensure those in rural areas or underserved communities with very few medical providers, can get to the right specialists in time. Mansi explains: “We’re not just detecting the stroke, but we're sending an alert to the neurosurgeon in that specialist hospital,” empowering doctors to decide much faster whether transferring the patient is required.  

Mansi co-founded Viz.ai in 2016 while in the Stanford MBA program. Eric Schmidt, former Google CEO, who was also his professor, decided to seed-fund the company. In 2018, Viz.ai’s stroke algorithm became the first of its kind to receive FDA approval. Since then, the company has continued to grow the types of data and list of conditions it can detect, boasting a total of 13 FDA-approved algorithms that can help triage a range of heart, lung and brain conditions.

Ray Kurzweil's eerily prescient predictions about AI are underpinned by a simple line chart. The chart, tracking the amount of computing power you could buy for a dollar over time, has grown exponentially for the past 85 years. Kurzweil initially used it to prioritize his inventions as a young programmer, focusing on a revolutionary print-to-speech machine for the blind in the 1970s, then early speech recognition software in the '80s. Eventually, he started writing about what he believed advances in computing power would bring and his predicting “took on a life of its own."

In 1990, Kurzweil correctly projected that AI would beat the best human player at chess before the turn of the millennium, and that mobile devices connected to a global information network would emerge in the decade that followed. In 1999, he forecast that by 2029, computers would match human intelligence in every domain. At the time, leading researchers like Geoffrey Hinton and Yoshua Bengio thought it would take much longer. They have since changed their tune.

OpenAI’s former chief scientist has had a tumultuous year. Ilya Sutskever, once widely regarded as perhaps the most brilliant mind at OpenAI, voted in his capacity as a board member last November to remove Sam Altman as CEO. The move was unsuccessful, in part because Sutskever reportedly bowed to pressure from his colleagues and reversed his vote. After those fateful events, Sutskever disappeared from OpenAI’s offices so noticeably that memes began circulating online asking what had happened to him. Finally, in May, Sutskever announced he had stepped down from the company.

On X (formerly Twitter), Sutskever praised OpenAI’s leadership and said he believed the company would build AGI safely. But his departure came at the same time as several other safety-focused staff left with more pessimistic public statements about OpenAI’s safety culture, which seemed to underline the impression that something fundamental had shifted at the company. The Superalignment team that Sutskever co-ran, which was aimed at developing methods to make advanced AIs controllable so they don’t wipe out humanity, had been promised 20% of OpenAI’s computing power to do its work. But the team sometimes struggled to access that compute and was increasingly “sailing against the wind,” Sutskever’s co-lead Jan Leike said in a series of messages announcing his own departure, in which he criticized the safety culture at OpenAI as having “taken a backseat” to building new “shiny products.” 

In June, Sutskever announced he was starting a new AI company, called Safe Superintelligence, which aims to build advanced AI outside of the market, to avoid becoming “stuck in a competitive rat race.” The implication that OpenAI is stuck in such a rat race is the most critical Sutskever has been of his former employer in public. 

Safe Superintelligence is at least the third AI company—after OpenAI and Anthropic—to be founded by industry insiders on the belief that they could build superintelligent AI more safely than their irresponsible competitors. But so far at least, with each new entrant, the race has only accelerated. Sutskever declined a request to be interviewed for this story via a spokesperson, who said he was in “what he describes as Monk Mode—heads down in the lab.”

Buy your copy of the TIME100 AI issue here

*Disclosure: OpenAI and TIME have a licensing and technology agreement that allows OpenAI to access TIME's archives.

Andrew Yao, widely-considered to be one of the most influential computer scientists of his generation, has left an indelible imprint on China’s approach to AI.

A Turing Award winner and the Dean of the Institute for Interdisciplinary Information Sciences at Tsinghua University, Yao has reportedly shaped some of China's biggest AI startups and a generation of academics. His students have gone on to found companies worth billions of dollars, like the software company Megvii and the autonomous vehicle company Pony.ai. Others have secured positions at top universities like Stanford and Princeton.

Since 2018, Yao has established four different AI Institutes across China, focusing on foundational research. Already in 2024, he has been appointed as head of Tsinghua’s newly-created College of AI, and was praised by President Xi Jinping for his unwavering dedication and remarkable achievements in teaching and research.
Yao has also become a leading voice in discussions of AI safety, vocal on the potentially catastrophic risks posed by future systems that might autonomously execute cyberattacks or assist in creating weapons of mass destruction.

In mid-November of 2023, Helen Toner made what will likely be the most pivotal decision of her career.

Together with three other members of OpenAI’s board, she voted to fire Sam Altman from his role as the company’s CEO. At the time, Toner and her fellow board members were silent about their reasons, saying only that Altman had “not been consistently candid” with them. In the information vacuum, a pressure campaign by Altman’s allies to reinstate him gained momentum. Silicon Valley luminaries, venture capitalists, and OpenAI’s biggest investor, Microsoft, joined the effort—as did most of OpenAI’s employees, whose equity in the company appeared to be at risk of losing most or all of its value. Five days later, Altman was back in the CEO’s chair. Outmaneuvered, Toner and all but one of the other board members who fired Altman agreed to step down. 

Amanda Askell’s nickname at Anthropic, the AI firm where she works, is the “Claude whisperer.”

Anthropic’s chatbot, Claude, has a reputation in the industry for being friendly, curious, and maybe a little more creative than its main rival ChatGPT. Askell is more responsible than anyone else for this thoughtfully engineered persona.

A philosopher by training, she leads the team at Anthropic that’s responsible for embedding Claude with certain personality traits and avoiding others. “It feels important to have a nuanced, rich conception of what it is to be good,” Askell says in an interview at Anthropic’s San Francisco headquarters. 

In an effort to help Claude be “good,” Askell has tuned it to openly admit to users when it’s unsure of its answer, to attempt to discuss ideas without bias, and to avoid both-sidesism when discussing settled issues like climate change. Most importantly, she’s engineered Claude to tell people it doesn't have feelings, memory, or self-awareness—that any personality that it might display is the product of complex language processing rather than evidence of an inner life.

In spite of all these measures, or perhaps because of them, there’s still something seemingly human-like about Claude’s personality. 

Some argue that making chatbots behave too anthropomorphically carries risks, including encouraging people to have inappropriate relationships with AI, or fostering inaccurate perceptions of how the technology actually works. But Askell is betting that in some scenarios, emulating human behavior can help users avoid falling into the dangerous trap of believing AI is all-knowing. “I was a bit worried about this idea that if you have this thing that feels robotic, that people might think of it as this authority,” she says, referring to the long-observed pattern of humans placing undue trust in machines. “The more you can signal that you’re talking with something that isn’t this grand source of authority on everything,” she says, it’s also more likely that people won't believe Claude’s outputs at face value. “It may feel more human-like—but that’s the line that you have to tread.”

*Disclosure: Investors in Anthropic include Salesforce, where TIME co-chair and owner Marc Benioff is CEO.