OpenAI Boosts Data Spending Alongside Tesla's Robotaxi Tests

The AI industry is experiencing a significant boom in specialized training data, with major labs like OpenAI and Anthropic spending over $10 billion annually. Public datasets no longer suffice, creating a lucrative market for human-annotated data. Companies such as Mercor and Handshake are key players, with Mercor, founded by Brendan Foody in 2023, pivoting to AI training data after a request from Scale AI for 1,200 engineers in 2024. This demand highlights the industrialization of AI, yet challenges persist, as Anthropic even paused model training due to data quality issues. Poor data quality remains a major hurdle for AI success. MIT research indicates that 95% of organizations see no measurable return on AI investments due to bad data and inadequate planning. Agentic AI, which can reason and act, critically depends on high-quality, "intelligent data" that is trusted and updated. Large Language Models (LLMs) can amplify misinformation if underlying data is messy. Despite these challenges, a Teneo survey reveals nearly 70% of CEOs plan to increase AI spending in 2026, though less than half report profitable AI projects currently. Successful AI adoption in 2025 often involved focusing on specific problems and "boring" back-end tasks, with employee involvement proving crucial, as seen when law firm Troutman Pepper Locke saved $200,000 using AI for attorney bios. Companies continue to integrate AI into their operations and products. Tesla, for instance, has begun testing driverless Robotaxis in Austin, Texas, without safety monitors, a first for the company. CEO Elon Musk and AI lead Ashok Elluswamy confirmed this step, noting Tesla's system relies solely on cameras, unlike Waymo's LiDAR. Meanwhile, Dropbox is bolstering its AI strategy by appointing Ross Tennenbaum as its new CFO, signaling increased investment in AI tools like Dropbox Dash, which integrates services such as Google Workspace and Slack, with an aim for profitability. In healthcare governance, Pacific AI received a 2025 InfoWorld Technology of the Year Award for its AI Policy Suite, a free framework that automates regulatory compliance by combining over 250 global laws. AI is also finding applications in unexpected areas, such as legal defense. Famed short-seller Andrew Left is using the AI chatbot Claude to prepare for his criminal trial in March, having uploaded case documents for analysis. While lawyers advise caution, AI can assist in understanding complex legal materials. On a broader societal level, discussions around the economic impact of AI continue. Andrew Yang advocates for Universal Basic Income (UBI) with a "Freedom Dividend" of $1,000 per month, but critics argue this amount is insufficient and a median income UBI would be prohibitively expensive, costing over $14 trillion annually, or 45% of the US GDP. This debate raises questions about massive inequality and the potential need to tax robot owners in an AI-driven economy.

Key Takeaways

• AI labs like OpenAI and Anthropic spend over $10 billion annually on specialized training data, creating a booming market for companies like Mercor and Handshake.
• Scale AI requested 1,200 engineers from Mercor in 2024, highlighting the high demand for human-annotated data in AI development.
• MIT research indicates 95% of organizations get no measurable return on AI due to bad data and poor planning, emphasizing the need for "intelligent data."
• A Teneo survey shows nearly 70% of CEOs plan to increase AI spending in 2026, but less than half report current AI projects are profitable.
• Tesla has initiated driverless Robotaxi tests in Austin, Texas, without safety monitors, a first for the company, relying solely on cameras for its self-driving system.
• Dropbox appointed Ross Tennenbaum as its new CFO to boost investment in AI tools like Dropbox Dash, aiming to turn them into profitable products.
• Pacific AI won a 2025 InfoWorld Technology of the Year Award for its AI Policy Suite, which automates regulatory compliance by integrating over 250 global AI laws.
• Short-seller Andrew Left is utilizing the AI chatbot Claude for legal defense preparation for his criminal trial, demonstrating AI's potential in legal analysis.
• Andrew Yang's proposed Universal Basic Income of $1,000 per month is deemed insufficient to address poverty in an AI-driven economy, with a median income UBI costing over $14 trillion annually.
• Successful AI adoption in 2025 focused on problem-solving and "boring" back-end tasks, with employee involvement, exemplified by Troutman Pepper Locke saving $200,000 using AI for attorney bios.

AI Training Data Creates New Billionaires and Big Business

AI labs now spend over $10 billion yearly on training data because public datasets are not enough. This has created a booming market for specialized, human-annotated data. Companies like Mercor and Handshake are key players, providing "grading rubrics" that need expert knowledge. Anthropic even paused training its model due to data problems. This trend shows AI is becoming industrialized, creating a profitable market for data providers.

Hidden Companies Profit Billions from AI Training Data

While big AI labs like OpenAI and Anthropic spend billions, lesser-known companies like Mercor and Handshake are making significant profits. Brendan Foody started Mercor in 2023 as a staffing agency. In 2024, Scale AI asked Mercor for 1,200 engineers, leading Foody to focus on AI training data directly. Labs now spend over $10 billion annually on specialized data, which is crucial for AI improvements. This data work, from simple image labeling to complex medical advice ratings, has become a highly profitable part of the AI industry.

Dropbox Appoints New CFO to Boost AI Product Revenue

Dropbox announced a change in its leadership this week. Timothy Regan, the long-time CFO, is stepping down. Ross Tennenbaum, who was president and former CFO of Avalara, will take over as the new CFO. This change comes as Dropbox plans to invest more in AI tools like Dropbox Dash. These tools connect services such as Google Workspace and Slack, and the company aims to turn them into profitable products.

Tesla Tests Driverless Robotaxis in Austin Texas

Tesla has started testing its Robotaxi rides in Austin, Texas, without any drivers inside the cars. This marks the first time Tesla has removed the safety monitor from its self-driving tests. Previously, these cars had a safety driver in cities like San Francisco, Oakland, San Jose, and Austin. Tesla CEO Elon Musk and AI lead Ashok Elluswamy confirmed this significant step. Unlike Waymo, which uses LiDAR, Tesla's self-driving system relies only on cameras.

Three Key Trends Shaped Company AI Use in 2025

In 2025, three main trends guided how companies used AI. Businesses found success by focusing on solving problems first, rather than just using AI for its own sake. Many companies saw great results by using AI for "boring" back-end tasks, which proved to be very effective. How companies involved their employees also played a big role in successful AI adoption. For example, law firm Troutman Pepper Locke saved $200,000 by using AI to redraft attorney bios during a merger.

New Survey Shows AI Enthusiasm Outpaces Real Results

A new global survey by Teneo shows that company leaders are very excited about artificial intelligence, but its performance is not always matching expectations. Nearly 70% of CEOs plan to spend more on AI in 2026. However, less than half of them say their current AI projects have made more money than they cost. AI has shown the best results in marketing and customer service, but it is harder to use in security, legal, and human resources. Many executives also believe AI will create more jobs, especially for entry-level and senior roles.

Pacific AI Wins Award for AI Policy Suite

On December 15, 2025, Pacific AI, a healthcare AI governance company, won a 2025 InfoWorld Technology of the Year Award. Their AI Policy Suite was recognized for its innovative way of automating regulatory compliance for AI. This free and always updated framework combines over 250 global laws and regulations. It helps organizations save many hours by simplifying complex AI rules and ensuring ethical and compliant AI development and use.

Universal Basic Income Fails to Solve AI Economy Problems

Andrew Yang continues to promote Universal Basic Income, or UBI, as a solution for job losses due to AI. He suggests a "Freedom Dividend" of $1,000 per month for every adult. However, this amount is not enough for a family to live above the poverty line. Providing a median income of $53,000 per year through UBI would cost over $14 trillion, which is 45% of the US GDP. The article argues that UBI, as currently proposed, does not address the massive inequality an AI-driven economy would create. It suggests that taxing robot owners might be necessary, raising new questions about power and control.

Bad Data Halts AI Progress in Boardrooms

Companies are investing millions in AI, but many are not seeing real progress or returns. MIT research shows that 95% of organizations get no measurable return on AI due to bad data and poor planning. Agentic AI, which can reason and act, needs high-quality data to work well. Data is often a company's most valuable asset but also its lowest quality resource. Large Language Models, or LLMs, can amplify bad information if the underlying data is messy or inconsistent. Companies that succeed with AI use "intelligent data," which is trusted, updated, and provides a complete view of critical information.

Short-Seller Andrew Left Uses AI Chatbot for Legal Defense

Famed short-seller Andrew Left is using the AI chatbot Claude to help prepare for his criminal trial in March. Left accidentally shared a chat log with Business Insider, showing him asking Claude questions about his legal defense. He uploaded a document related to his case and asked the AI for a quick analysis. Left faces federal charges for allegedly manipulating stock prices and making false statements, which he denies. While lawyers caution against over-reliance, AI can be a valuable tool for understanding complex legal materials.

Sources

• Meet the Billionaires Selling AI Its Training Data
• Who’s making the most money in AI? It’s not who you think
• Should Dropbox’s New CFO and AI Focus Require Action From Dropbox (DBX) Investors?
• Tesla Begins Testing Robotaxi Rides In Austin With No One Inside The Car
• The 3 trends that dominated companies’ AI rollouts in 2025
• Survey finds a growing gap between AI enthusiasm and performance
• Pacific AI Named a 2025 InfoWorld Technology of the Year Award Winner for the AI Policy Suite, Automating Regulatory Compliance
• Why universal basic income still can’t meet the challenges of an AI economy
• This is the AI question that stops a boardroom cold
• How famed short-seller Andrew Left has been using AI to prepare for his criminal case