In the age of artificial intelligence, ‘monitoring the monitor’ cannot be left to a single actor. The issue needs to be viewed through the lens of a constantly re-establishing balance between technological progress, democracy, and ethics.

Who will monitor those tasked with protecting society and maintaining order if they abuse their power? While the answer to this age-old question is often found in concepts like the rule of law, separation of powers, independent judiciary, regulatory and supervisory bodies, a free press, and freedom of expression, this answer is not always satisfactory. In the age of artificial intelligence (AI), this question needs to be revisited. Because AI is unlike other technologies.

In recent months, there have been very rapid developments in AI models, and the potential disruptions these models could cause at the economic, social, political, cultural, intellectual, and even global levels are becoming more apparent.

In this respect, Anthropic's new model, Mythos, is very noteworthy. Claimed to have the ability to detect vulnerabilities in operating systems and browsers, this model has been made available only to a limited number of partners instead of being released for general use.

The following questions need to be answered:

● Will companies or governments control technologies that could have dire impacts on the lives of societies?

● Who will oversee the large technology companies developing these models?

● Who will oversee the governments using these AI models?

Simply answering these questions with "external oversight" is insufficient. An interesting answer comes from technicians working at large technology companies. Researchers, engineers, and security teams working at these companies are the ones most aware of the risks of the systems they develop. We are seeing that those who build these systems can no longer remain silent.

Google's planned AI deal with the Pentagon has been protested by over 560 Google employees. In an open letter to the CEO, the employees called on the US government not to allow Google to use its AI technology for covert military operations.

"We want artificial intelligence to benefit humanity, not be used in inhumane or extremely harmful ways," the Google employees stated, adding, "We believe our affinity with this technology creates a responsibility to highlight and prevent its unethical and dangerous uses." Similar objections have come from other companies. At the end of February, OpenAI's agreement with the Pentagon faced backlash from its own employees.

Anthropic CEO Amodei also stated that his company would not allow its AI models to be used for lethal autonomous weapons and mass surveillance.

International organizations have also warned that AI capabilities are developing faster than governance. Organizations such as the EU, UN, OECD, World Bank, and WEF have published reports highlighting the security risks of AI, its potential to deepen inequalities, and the need to develop governance rules to protect fundamental rights and democracy.

The idea of ​​adhering to ethical principles is good, but how is this possible under a capitalist system? The Palantir manifesto, which shows the convergence of the capitalist company's profit motive and the capitalist state's ambition to dominate the world, raises this question.

Ultimately, AI companies are also subject to the rules of the market economy: they must aim for high profits. They must also work in harmony with their biggest customer, the capitalist state.

Anthropic CEO Amodei emphasizes another point: AI can become a terrifying tool in the hands of authoritarian states. Therefore, democratic states must necessarily be ahead of authoritarian states in AI technologies. However, the use of AI models developed to combat authoritarianism by democratic countries risks transforming them into authoritarian surveillance societies.

This brings us back to the initial problem: a collective action problem that is far from easy to solve.

If companies and countries that do not adhere to ethical and security standards can use AI technologies to disable those that do, who should ensure effective surveillance, and how?

The answer lies not in limiting technological development, but in determining the institutional and geopolitical framework within which that development takes place. In the age of artificial intelligence, ‘monitoring the monitor’ cannot be left to a single actor. The issue must be viewed through the lens of a constantly re-establishing balance between technological progress, democracy, and ethics.

The components of the mechanism to establish this balance are beginning to emerge. The objections of researchers in AI companies, regulatory initiatives such as the EU's AI Law, and proposals for an international oversight system similar to the Atomic Energy Agency are all signs of this. None of them are sufficient on their own, but together they point in one direction: this oversight must be carried out not by a single actor, but by an ecosystem of many actors—technicians, civil society, democratic institutions, and international treaties—balancing each other.