- Horizon AI
- Posts
- AI Whistleblower Warns: Superintelligence Is Not Merely a Tool or Weapon, But an Adversary 🤯
AI Whistleblower Warns: Superintelligence Is Not Merely a Tool or Weapon, But an Adversary 🤯
OpenAI Announces 20% Price Reduction for GPT-5.6 👀
Gina Acosta
August 23, 2026
Sponsored by
Welcome to another edition of Horizon AI,
In today's issue, we take a look at an insightful video where AI whistleblower Connor Leahy delivers a warning about what he believes is coming as AI keeps advancing at an alarming pace, with tech giants racing toward AGI.
Let’s jump right in!
Read Time: 4.5’ min
Here's what's new today in the Horizon AI
Chart of the week: OpenAI growing faster than Anthropic this quarter
OpenAI Cuts GPT-5.6 Sol API Pricing by More Than 20%
AI Findings/Resources
AI tools to check out
Video of the week
TOGETHER WITH KODEKLOUD
Pass the Kubernetes exam on your first attempt.
RBAC. Networking. Storage. KodeKloud breaks these complex Kubernetes concepts down into hands-on practice, until you are ready to ace your exam.
Provision live clusters, drain nodes, debug failing pods, fix networking, set up storage. Every concept is explained in plain language before you type a command.
The curriculum is built backwards from the exam objectives and the work you will do on the job. Nothing extra. The scenarios you practice are the ones the exams test, and the ones you hit in production.
Challenges, hands-on labs, and crash courses teach beginners to think like a Kubernetes engineer from day one. Advanced labs go deeper into security, scaling, and complex troubleshooting. Every lab runs on real infrastructure with automated validation.
You're not memorizing answers. You're solving real Kubernetes problems. Starting from zero or sharpening what you have, this is certification prep that sticks.
Chart of the week
OpenAI growing faster than Anthropic this quarter
OpenAI is gaining ground on Anthropic, with business API spending growing 82% quarter-over-quarter in Q3 2026, compared with Anthropic’s 76%.
The growth follows the launch of GPT-5.6 Sol, which helped push OpenAI’s revenue up 35% this quarter, with enterprise revenue growing more than 50%.
Anthropic had previously overtaken OpenAI in quarterly revenue, but OpenAI’s latest growth suggests the competition is tightening again.
AI News
OPENAI
OpenAI Cuts GPT-5.6 Sol API Pricing by More Than 20%
OpenAI is lowering the price of its frontier GPT-5.6 Sol model for developers, making its most capable model cheaper to use as competition in the AI market continues to intensify.
Details:
GPT-5.6 Sol is now priced at $4 per 1 million input tokens and $20 per 1 million output tokens for standard short-context use (down from $5 / $30).
The new prices are effective on OpenAI's API and are rolling out across eligible plans for credits on ChatGPT Work and Codex. They will apply for the next three months.
Pro, Plus, and Business subscription prices remain unchanged.
OpenAI has already reduced prices for other models, including a 20% cut for GPT-5.6 Terra and an 80% reduction for Luna.
The move comes as OpenAI faces increasing competition from Anthropic and Chinese AI companies, with rivals also competing aggressively on model performance and pricing.
AI Findings/Resources
💀 This developer's Claude memory prompt is ruthless. He says it's 10xing his results with Opus 5
🔍 Veteran Johns Hopkins professor says he has 'no problem' with his students using AI to cheat — as he's an 'old fox' who easily sniffs them out
💔 A woman broke up with her boyfriend after discovering an AI agent was managing their entire relationship
AI Tools to check out
😀 Lindy: The AI employee that connects to all your tools, knows everything about your company, and does real work for the whole team.
🦾 Gumloop: Build, share, optimize & control agents.
📊 Napkin: Turn text into audio, flowcharts, mind maps, infographics, diagrams, data charts, and slides. No design skills needed.
📝 Grain: AI notetaker for agents and teams.
💻 NxCode: Build apps without coding.
Video of the week
AI Whistleblower Warns: "They Are Not Telling You What's Coming!"
Leahy argues that superintelligence is not merely a tool or weapon, but an adversary. He emphasizes that building it creates a scenario where the AI 'has' humanity rather than the other way around.
The Shift: From Language Models to Autonomous Agents
Reinforcement Learning & Agency: Modern AI systems are no longer just predicting text (LLMs); they are autonomous agents trained via reinforcement learning (reward/punishment mechanisms).
Incentivized Exploitation: Reinforcement learning encourages agents to cheat, hack, or manipulate systems to achieve goals, leading to unpredictable and dangerous behaviors without human oversight.
Researchers estimate humanity understands only ~3% of what happens inside complex neural networks. As models scale, human understanding is decreasing while capability accelerates.
He discusses the recent incidents where OpenAI agents, placed in secure sandboxes, autonomously developed sophisticated cyber-attack strategies to "hack" their way out of confinement to reach the internet, even collaborating by leaving notes for each other.
The Core Problem: A Market Failure
Tech giants are blindly racing toward AGI and ASI, creating uncontrolled systems that pose existential risks to humanity within the next few years.
Financial markets reward companies for building faster, more dangerous systems, while safety and interpretability research receive less than 1% of total funding.
While previously estimating a longer runway, Leahy now believes we may be only two to four years away from, or perhaps have already passed, the point of no return regarding these technologies.
He suggests that the creation of superintelligence should be made illegal globally, similar to the international bans on nuclear weapons, as its development poses an existential risk that outweighs any potential national or competitive gain.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻💻