VCs are selling shares of hot AI companies like Anthropic and xAI to small investors

Venture capitalists are keen to invest in leading AI companies like Anthropic and xAI, but often struggle to secure positions on their cap tables. Small investors, including family offices and high-net-worth individuals, are accessing shares of these top private startups through special purpose vehicles (SPVs). SPVs allow multiple investors to pool their resources to share an allocation of a company’s stock, often facilitated by those with direct access to the shares. These vehicles, while not new, are increasingly being used to secure shares in prominent AI firms. The investors behind SPVs typically charge significant fees and retain a portion of the profits.

SPVs are becoming more common as early backers of popular AI startups exercise their pro-rata rights, which enable them to purchase additional shares during fundraising rounds. When early investors cannot afford more shares, they form SPVs to raise the necessary capital from other investors, often charging extra fees. This allows smaller investors to gain access to high-demand companies that would otherwise be out of reach. Some venture capitalists offer SPV access to their limited partners or use brokers to reach a broader pool of potential investors, resulting in multiple SPVs on a single company’s cap table, each with varying terms.

The SPV market can be complex and inconsistent, with fees and profit-sharing arrangements varying widely. Ken Sawyer, co-founder of Saints Capital, notes that SPV sponsors may charge up to 2% of the invested capital and retain 20% of the profits. Additionally, some SPVs are layered on top of others, further increasing costs. For example, Menlo Ventures raised a $750 million SPV for Anthropic, and some funds that invested in it resold portions of their allocation through secondary SPVs, adding extra fees. This layered approach adds to the complexity and cost for smaller investors looking to participate in these high-profile AI startups.

Hollywood Nightmare? New Streaming Service Lets Viewers Create Their Own Shows Using AI

Fable Studio has launched Showrunner, a new streaming platform that allows users to create their own animated content using AI technology. This platform, announced on Thursday, can write, voice, and animate episodes based on user prompts, offering extensive control over dialogue, characters, and shot types. Users can join a waitlist for the free testing version, which will last until the end of the year. The platform currently features 10 AI-generated animated shows of various genres, including Exit Valley, Ikiru Shinu, and Sim Francisco. Showrunner's tech, which currently supports only animation, enables users to create scenes that can be stitched into full-length episodes, with the best user-created episodes potentially included in the series catalog and earning revenue.

Fable's vision, described by CEO Edward Saatchi as becoming the "Netflix of AI," includes the capability for users to generate new episodes of their favorite shows on demand. This development, however, raises concerns within the creative industry about the potential displacement of human labor by AI tools. The company faced criticism and intrigue when it released an AI-generated episode of South Park, showcasing both the potential and the comedic limitations of AI-generated content. Despite mixed reactions, this effort underscored the evolving role of AI in content creation and its potential impact on traditional production methods.

Showrunner, powered by Fable's proprietary AI model developed from open-source systems like OpenAI and Stable Diffusion, underscores the controversial use of publicly available data to train AI. Saatchi emphasizes that the originality of the content is crucial to the platform's success. Fable's approach to involving users in content creation not only aims to reduce production costs but also highlights the growing significance of union protections against AI's encroachment into the entertainment industry. The success of Showrunner could signal a shift in how content is produced and consumed, blending traditional storytelling with innovative AI capabilities.

Saudi fund invests in China in an effort to create a rival to OpenAI

A Saudi Arabian fund, Prosperity7, which is part of Aramco's venture capital arm, has become the first foreign investor in China's leading generative AI start-up, Zhipu AI. The investment, part of a $400 million funding round, values Zhipu AI at around $3 billion. This move signifies Saudi Arabia's intent to support AI ecosystems that counterbalance US dominance. Other leading Chinese AI start-ups, such as Moonshot AI, MiniMax, and 01·ai, have so far relied on domestic funding due to US restrictions.

The Saudi investment is part of a broader trend of collaboration between Saudi state investors and Chinese technology companies. For instance, Lenovo recently issued $2 billion in convertible bonds to Alat, a subsidiary of Saudi Arabia's Public Investment Fund, in exchange for setting up its regional headquarters and a manufacturing plant in Riyadh. This aligns with Saudi Arabia's strategy to attach stringent requirements to their tech investments. Meanwhile, US pressures on global tech investments in China have been intensifying, with companies like Abu Dhabi's G42 divesting from Chinese tech under US influence.

Zhipu AI, the largest Chinese generative AI start-up by staff, has over 800 employees and is headquartered in Beijing's Haidian district. The company, previously funded by Alibaba Cloud, Tencent, and Meituan, sells an "AI-in-a-box" product that allows companies to use its large language model on their premises for enhanced data protection. Founder Tang Jie, a former Tsinghua University professor, recently appeared at a Prosperity7 investment forum in Beijing, showcasing the growing ties between Saudi investors and Chinese AI firms.

Hugging Face says it detected ‘unauthorized access’ to its AI model hosting platform

Late Friday afternoon, AI startup Hugging Face disclosed an “unauthorized access” incident involving its Spaces platform, used for creating, sharing, and hosting AI models and resources. The breach was related to Spaces secrets, which are private pieces of information like account keys and tokens. As a precaution, Hugging Face has revoked several tokens and advised users to refresh their keys and consider switching to more secure, fine-grained access tokens. The extent of the impact remains unclear, and the company is collaborating with cybersecurity forensic specialists and law enforcement to investigate.

Hugging Face's spokesperson mentioned a significant increase in cyberattacks recently, likely due to the platform’s growing popularity and the mainstream adoption of AI. This incident is the latest in a series of security challenges for Hugging Face. In April, a vulnerability discovered by cloud security firm Wiz, and later fixed, allowed attackers to execute arbitrary code. Earlier, security firm JFrog found code uploaded to Hugging Face that installed malware, and HiddenLayer identified potential abuses of Hugging Face's Safetensors format.

In response to these security issues, Hugging Face announced a partnership with Wiz to use their vulnerability scanning and cloud environment configuration tools, aiming to bolster security across the platform and the broader AI/ML ecosystem. Hugging Face emphasized its commitment to improving security and mitigating the inconvenience caused by such incidents, using this breach as an opportunity to strengthen its infrastructure.

Other stuff

• Nvidia’s G-Assist is an AI chatbot that guides you through games and optimizes your PC

• Computer makers unveil Nvidia Blackwell systems for AI rollouts

• The rise of ChatCCP

• Voice cloning of political figures is still easy as pie

• LLMs Aren’t Just “Trained On the Internet” Anymore

• What We Learned from a Year of Building with LLMs (Part II)

• AI training data has a price tag that only Big Tech can afford

• Microsoft Asked For A Cut From Future OpenAI Profits As Apple Plans To Integrate AI Into iOS 18

• Corporations looking at gen AI as a productivity tool are making a mistake

• Winnipeg man caught in scam after AI told him fake Facebook customer support number was legitimate

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.