Masak warns that as argument becomes more dangerous than calculation the model race shifts to persuasion
Elon Musk admitted what senior technology leaders have been repeating in their statements: the ability of superintelligent AI to persuade may break human mental defenses without exception. In a post on X, the owner of X.AI warned that a very large number of people will suddenly be defeated by a highly persuasive superintelligence, openly acknowledging that he may personally be one of those who fall under this influence.
This statement shifts the focus of technical debate from traditional concerns about machines’ superiority in calculationThis statement shifts the focus of technical debate from traditional concerns about machines’ superiority in calculation, analysis, and task execution to a more sensitive domain that directly affects beliefs and daily decision-making. The seriousness of this trajectory increases as intelligent systems develop the ability to understand human behavior and craft highly refined personalized messages. The core problem lies in the fact that a system capable of producing coherent and persuasive arguments may cause users to place their trust in inaccurate information or adopt choices that harm their immediate interests, whether in financial, political, or social spheres. Nevertheless, the facts confirm that persuasive skill does not imply that a machine possesses self-awareness or independent intentions; the actual impact remains tied to the nature of available capabilities, usage patterns, and the level of stringent controls imposed on it.
The system's ability to craft a convincing argument does not mean it possesses consciousness, but it gives it influence that guides human decisions without their awareness.
This warning comes at a time when polarization within the technology sector is intensifying between calls for caution and the pressure of frenzied competition. Dario Amodei, the chief executive of Anthropic, has explicitly called for slowing the development of the most powerful AI models, warning of the possible consequences if the pace of development exceeds the ability of companies and governments to ensure the safety of these systems. However, this stance collides with a rapid race that major companies are pursuing to develop models with greater logical reasoning, complex task execution, and direct interaction with users, placing those entities before a tough equation that balances technical leaps with building effective tools to test, monitor, and limit the misuse of models.
On the regulatory front, major U.S. companies including Anthropic, OpenAI, Google and Meta have signed an agreement with the White House that outlines voluntary safety measures. However, relying on the voluntary nature of these commitments leaves major questions about the absence of field-verification mechanisms for their implementation and about whether they are sufficient to address rapidly accelerating technological risks that do not wait for slow approvals.
This shift in the concept of risk rearranges the priorities of institutions in the Gulf, Egypt and the Arab East in a practical and direct way. When banks, investment firms and government bodies in the region rely on artificial-intelligence models to analyse financial policies or guide administrative decisions, the main dilemma becomes the lack of strict safeguards against the persuasive power of automation. The risk of misleading inference does not appear as a digital error that can be detected, but rather seeps in through coherent logical arguments that lead investment committees or planners to uncalculated decisions. Regional business environments cannot fall back on voluntary commitments made in Washington without independent audit mechanisms. The situation here calls for the establishment of local oversight committees that examine the pathways of automated reasoning and test its outputs before they are delegated.
Technical safety is not achieved through voluntary commitments, but by building internal inspection systems that weigh the argument on the scale of reality.
These data confirm that the gap between information and a convincing argument has become the dividing line between safety and institutional stumbling. And if Masak himself, who leads the development of competing models, acknowledges the possibility of being subject to super-intelligent persuasion, then your first step in leading any technical team is to abandon excessive confidence in the machine’s rhetoric and build a human audit trail that holds the model accountable for factual accuracy rather than persuasive skill.