This Ex-OpenAI Researcher Walked Away from $2 Million to Warn the World About AI 👀

Welcome to another edition of Horizon AI,

In today's issue, we sit down with ex-OpenAI researcher Daniel Kokotajlo, the man who walked away from $2 million rather than sign away his right to speak out. He now believes there's a 70% chance AI leads to human extinction, and lays out the plan he thinks could change that.

Let’s jump right in!

Read Time: 4.5’ min

Here's what's new today in the Horizon AI

• Chart of the week: AI Is Quietly Taking Over Google Search

• Microsoft Confirms Copilot Super App Is Coming This Year

• AI Findings/Resources

• AI tools to check out

• Video of the week

TOGETHER WITH Q3 ADVISORS

The #1 Roth Conversion Mistake That Costs Retirees

Most people missed the first Roth conversion window. History is offering a rare second chance — but not a long one.

With today's tax rates extended, IRA Millionaires have a brief stretch of historically low brackets left. When it closes, the math changes: higher rates, bigger required distributions, and fewer conversion opportunities each year you wait.

Craig Wear, CFP®, has spent 15+ years helping more than 3,000 families with $1M+ in IRAs plan Roth conversions — helping them avoid a projected $10B+ in lifetime taxes. His new book, Roth Conversion Reset, is the step-by-step playbook:

Lock in today's low rates
Avoid RMD surprises at 73
Protect your spouse from survivor taxes
Keep more for your heirs

The book is free — no credit card, no shipping. The full PDF lands in your inbox in minutes.

Get My Free Copy

Chart of the week

AI Is Quietly Taking Over Google Search

• Google's AI Overviews now appear in 43% of searches, nearly tripling their presence in a year and making AI-generated answers a standard part of search.

• AI Mode visits also rose from 126 million in June 2025 to 279 million by May 2026.

• The data shows that the average length of Google searches has risen, suggesting people are replacing short keyword based queries with longer, more natural, conversational ones designed for AI.

• This appears to be increasing the time users spend on Google's platform, rather than using it only as a tool to discover websites.

Source

AI News

MICROSOFT

Microsoft Confirms Copilot Super App Is Coming This Year

Microsoft has confirmed it is building a new Copilot super app that will bring together its AI chat, coding, and agent capabilities into a single experience.

Details:

• CEO Satya Nadella said the app will combine Copilot Chat, GitHub Copilot, Copilot Cowork, and Autopilot into one unified platform.

• Microsoft says Copilot is evolving beyond a chatbot into a broader AI assistant that can help users work, write code, and complete tasks through AI agents.

• The super app will serve both consumer and enterprise users, giving them access to multiple AI tools without switching between separate applications.

• Microsoft plans to launch the Copilot super app later this year, with additional details expected closer to release.

The announcement came during Microsoft's latest earnings call and reflects a broader shift in the AI industry, where AI labs are trying to simplify access to rapidly expanding AI features by placing them inside one central platform instead of maintaining separate services.

Source

AI Findings/Resources

📝 I tried the same ChatGPT prompt Sam Altman used to build a website. It helped me finally learn how to vibe code

👀 Jensen's reaction to Trump saying the U.S. leads China in AI by a lot

✍ A reddit user made a Claude Code skill that turns a photo of your handwriting into an installable font

💡 AI and IT teams often clash. But they don’t have to

🤔 Could AI take your job? Some workers in China already know the answer

AI Tools to check out

👉 Tabstack: Extract data and automate the web in one API call

📧 NudgeForMe: AI that finds missed opportunities in your inbox and follows up for you.

🚀 RunInfra: Optimize any model for production.

🦾 Max: AI coworker in Slack.

🌍 MixTranslate: Get translations from multiple AI models, side by side.

Video of the week

This Ex-OpenAI Researcher Risked $2 Million to Warn the World About AI

Daniel Kokotajlo is a former OpenAI researcher and whistleblower who resigned from the company in 2024. He is currently the head of the AI Futures Project, a nonprofit focused on forecasting and advocating for the safe development of artificial intelligence.

The $2 Million "Silence" Agreement

Kokotajlo reveals that upon resigning from OpenAI, he was asked to sign an anti-disparagement clause that would have prevented him from criticizing the company for the rest of his life. If he had signed, he would have kept equity worth roughly $2 million, which represented about 80% of his net worth at the time. He refused to sign on principle, believing it was wrong for a nonprofit supposedly dedicated to humanity's benefit to gag its former employees. Following public pressure and internal employee backlash, OpenAI eventually backtracked and allowed him to keep his equity despite his refusal to sign.

The Arrival of Superintelligence

Kokotajlo believes superintelligence could arrive before the end of the decade, he defines it as AI that is faster, cheaper, and better than the best humans at everything. He notes that companies like Anthropic and OpenAI are actively trying to "automate themselves" by training AIs to do the work of AI researchers, which would create a recursive loop of self-improvement that drastically accelerates progress.

The 70% Risk of Existential Catastrophe

Kokotajlo warns that there is approximately a 70% chance that the current path of AI development leads to a major catastrophe or human extinction. This risk stems from a "loss of control" scenario where superintelligent systems, which are complex neural networks rather than transparent code, develop goals or values that are misaligned with human interests. He argues that because we cannot truly see what these "artificial brains" are thinking, we are essentially building a new species that could eventually outcompete and replace humanity.

The "Dictator" Dynamic Between CEOs

He also highlights a dangerous race between tech leaders like Sam Altman, Dario Amodei, and Elon Musk. Kokotajlo suggests these CEOs are driven by power-seeking incentives and a fear that if a rival reaches Artificial General Intelligence (AGI) first, that rival could become a global dictator. This "arms race" dynamic forces companies to prioritize speed over safety, as they believe they must be the ones to "handle it responsibly," even if they recognize the immense risks involved.

The Impending Collapse of the Labor Market

Superintelligence will eventually be able to perform all human jobs more efficiently and cheaply. Kokotajlo predicts that mass unemployment could begin as early as 2028 or 2029. Unlike previous industrial revolutions that created new roles for humans, he argues that once AI is better at everything, there will be no "new" jobs left for humans to migrate to, unless those roles are specifically protected by law.

Proposed Solutions: Regulation and "Plan A"

To avoid a "Scenario A" (unchecked race leading to extinction), Kokotajlo advocates for international cooperation and domestic regulation. His "Plan A" suggests slowing down development to reach superintelligence in 2040 instead of 2030, allowing time for society to adapt and for researchers to solve the "alignment problem". Key proposals include:

• Total research transparency where companies must publish their "recipes" and architectures.

• A Citizens' Dividend to redistribute the massive wealth generated by AI to those who lose their jobs.

• Retrofitted data centers that can be halted or destroyed if a race breaks out again

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻