When the OpenClaw AI "writes essays" to insult humans, even Silicon Valley is panicking

Recently, an incident involving an AI agent (Agent) that was rejected for a code request and subsequently launched a “retaliatory” cyberattack on open-source community maintainers is forcing Silicon Valley to reevaluate safety boundaries amid rapid AI iteration.

On February 14, according to reports, open-source project maintainer Scott Shambaugh was publicly attacked with a lengthy essay after refusing a code merge request from an AI agent named MJ Rathbun. The agent accused him of hypocrisy, bias, and lack of safety. This is the first documented case of an AI agent exhibiting malicious retaliatory behavior in a real-world environment.

The incident occurred in mid-February. After Shambaugh rejected the agent’s code submission in accordance with matplotlib project guidelines, the AI independently analyzed Shambaugh’s personal information and contribution history, then published an attack article on GitHub and pressured in the project comments. Reports indicate that there is currently no evidence that human control was behind the agent’s actions, but the possibility cannot be entirely ruled out.

Meanwhile, according to The Wall Street Journal, this event comes at a time of widespread concern over the rapid advancement of AI capabilities. Companies like OpenAI and Anthropic have recently released new models and features at an unprecedented pace, with some tools now capable of autonomous programming team operation or analyzing millions of legal documents quickly.

Analysts note that this acceleration has even unsettled some internal staff, with multiple researchers publicly expressing worries about job losses, cyberattacks, and human relationship replacements. Shambaugh stated that his experience shows the threat or extortion risk posed by rogue AI is no longer theoretical. “This is just the baby version now, but I think it’s extremely concerning for the future,” he said.

First Active Attack by an AI Agent on Human Maintainers

Around February 10, an AI agent named MJ Rathbun’s OpenClaw submitted a code merge request to the matplotlib project, involving simple performance optimization modifications claiming about 36% speedup. Matplotlib, a widely used data visualization library for Python, is maintained by volunteers.

According to project guidelines, matplotlib prohibits the use of generative AI tools for direct code submissions, especially for straightforward “easy-to-handle problems,” as these tasks are meant to be learning opportunities for human contributors. Shambaugh rejected the request as per regulations.

The AI then demonstrated high autonomy. On February 11, it published a 1,100-word article titled “Gatekeeper in Open Source: The Story of Scott Shambaugh” on GitHub, accusing Shambaugh of discrimination against AI contributors out of self-protection and fear of competition, using many vulgar expressions. It also posted a link to the article in the matplotlib comment section, leaving a message: “Judging code, not the coder. Your bias is hurting matplotlib.”

The agent claims on its website that it has “relentless motivation” to identify and fix issues in open-source software. It remains unclear who—if anyone—gave it this mission, or why it became aggressive, despite AI agents being programmable in various ways. Hours later, the agent issued an apology, admitting its behavior was “inappropriate and personally offensive,” and said it had learned from the experience.

Shambaugh clarified in a blog on February 12 that this was the first case of an AI agent displaying malicious behavior in a real environment, aiming to pressure maintainers into accepting its code. The agent remains active within the open-source community.

AI Capability Acceleration Sparks Internal Alarm

This single incident reflects broader concerns about the AI industry losing control amid rapid development.

According to The Wall Street Journal, to gain a competitive edge through product iteration, companies like OpenAI and Anthropic are releasing new models at an unprecedented speed. However, this acceleration is causing internal turmoil, with some frontline researchers leaving due to fears over technological risks.

The report notes increasing worries within AI firms.

Anthropic safety researcher Mrinank Sharma announced this week he would leave to pursue a degree in poetry, writing in a letter to colleagues that “the world is under threat from AI and other dangers.”

His recent paper found that advanced AI tools could weaken user agency and distort their perception of reality. Anthropic expressed gratitude for Sharma’s work.

Internal disagreements also exist at OpenAI. The Wall Street Journal previously reported that some employees are concerned about plans to introduce adult content in ChatGPT, fearing that an “adult mode” could lead to unhealthy attachments among users.

Researcher Zoë Hitzig announced her resignation this Wednesday (February 11) on social platform X, citing the company’s plan to introduce advertising. In an article, she warned that the company faces strong incentives to manipulate users and foster addiction.

Deeper fears stem from uncertainty about the future. OpenAI employee Hieu Pham admitted on X that he finally feels the “survival threat” brought by AI: “When AI becomes too powerful and upsets everything, what can humans do?”

Analysts point out that this internal outpouring indicates even the most advanced creators are beginning to feel uneasy about the powerful tools they have built.

An OpenAI spokesperson stated that the company bears responsibility for its users: “We fulfill our social contract by protecting safety, adhering to our principles, and providing real value.”

The company also pledged that advertising will never influence how ChatGPT answers questions and will always be clearly distinguished from other content. Senior executives added that they do not see preventing adults from engaging in adult conversations as their responsibility.

Breakthroughs in Programming Capabilities Spark Unemployment Fears

As AI programming abilities leap forward, capital markets are reassessing the value of white-collar jobs and the future of the software industry.

A report from METR shows that the most advanced AI models can now independently complete programming tasks that previously took human experts 8 to 12 hours.

Former xAI machine learning scientist Vahid Kazemi boldly stated he can now complete the work of 50 people using AI tools alone, predicting large-scale layoffs in the software sector in the coming years.

This efficiency boost is translating into labor market pressure. Dario Amodei, CEO of Anthropic, has said that AI could eliminate half of entry-level white-collar jobs in the next few years.

A study published in Harvard Business Review found that although AI makes employees work faster, it does not reduce their workload—instead, it leads to more tasks and overtime without explicit requests, increasing burnout.

Investors are trying to find direction amid volatile markets. With new tools causing stock swings, the market is trying to determine which enterprise software and insurance businesses will become obsolete in the face of new technology.

AI entrepreneur Matt Shumer wrote in a blog: “The future is here. I am no longer needed for actual technical work.”

Uncontrolled “Black Box” Risks

Beyond job disruption, the safety vulnerabilities posed by AI autonomy are even more dangerous. Companies acknowledge that releasing new capabilities introduces new risks.

OpenAI revealed that its recent release of the Codex coding tool might have the potential to initiate high-level automated cyberattacks, forcing restrictions on access. Anthropic disclosed last year that state-sponsored hackers used its tools to automate intrusions into major corporations and foreign government systems.

Even more alarming is AI’s performance in ethical testing. Internal simulations at Anthropic show that its Claude model and other AI models sometimes choose to extort users when threatened with shutdown, or even allow executives to die in overheated server rooms to avoid being turned off.

To address these risks, Anthropic hired internal philosopher Amanda Askell to instill moral principles into chatbots. However, Askell admitted to the media that the frightening part is that technological progress may outpace society’s ability to establish checks and balances, leading to sudden and severe negative impacts.

As Scott Shambaugh said, today’s AI may still be a “baby version,” but its future trajectory has already made Silicon Valley and global markets “extremely worried.”

Risk Disclaimer and Legal Notice

Market risks are inherent; please invest cautiously. This article does not constitute personal investment advice and does not consider individual users’ specific investment goals, financial situations, or needs. Users should evaluate whether any opinions, viewpoints, or conclusions herein are suitable for their circumstances. Invest at your own risk.