While the global debate is focused on artificial intelligence, in terms of issues of functionality, privacy and ethics, new models have emerged capable of discovering unknown vulnerabilities in software, which has contributed to moving the concern to a more sensitive level, which is the security of countries’ digital infrastructure.
The Mythos model, developed by the American company Anthropic, has, within a few weeks, become a headline in both Western and Chinese newspapers, as an example of the power of artificial intelligence in detecting vulnerabilities, the risks this raises, and the opportunities it creates for the cybersecurity industry, according to specialists.
Read also
list of 2 itemsend of list
One model occupies the world
On April 7, Anthropic announced its new model, “Cloud Mythos Preview,” which is an advanced artificial intelligence model specifically designed to discover and address security vulnerabilities in software, with extensive capabilities for automated code analysis.
The announcement “registered an unprecedented response among policymakers and regulators due to its strong ability to identify and exploit cyber vulnerabilities in critical software,” the South China Morning Post said in an editorial from a series on the Mythos model issue.
The US National Security Agency and other US government agencies have already begun using the model, after Anthropic made the Mythos model available to a group of US companies, including Cisco, JP Morgan Chase, and Nvidia, for use in securing their sensitive software as part of an initiative known as “Project Glasswing.”
The newspaper noted that Chinese companies were not invited to use the model to fill their cyber vulnerabilities, in light of China’s classification as an “enemy state” by the company, and with Anthropic’s services continuing to be banned in the Chinese mainland and Hong Kong.
A new weapon in cyberspace
The Global Times, in its Chinese edition, discussed the issue of the “Mythos” model and quoted the “Financial Times” newspaper as saying that the model is capable of discovering software defects faster than traditional human methods, and that it has crossed the threshold of major capabilities, as it can automatically discover and verify unknown security vulnerabilities in common operating systems and browsers.
The model discovered thousands of high-risk vulnerabilities, some of which had been around for 10 or even 20 years without anyone noticing
The newspaper’s report also indicated that the model discovered thousands of high-risk vulnerabilities, some of which had existed for 10 or even 20 years without anyone noticing, and that electronic attacks supported by artificial intelligence are expected to rise by 89% in 2025 compared to the same period in 2024.
Experts surveyed by the newspaper describe this ability as a “double-edged sword”, as it can be used to plug vulnerabilities efficiently and quickly, but its misuse will significantly reduce the value of technical barriers to electronic attacks, which increases the risks of systems being hacked, data leaked, and critical infrastructure paralyzed.
Global Times quotes Christina Cacioppo, CEO of the security and compliance platform Vanta, as saying, “The pace and complexity of cyber attacks supported by artificial intelligence big data models has actually increased, and most companies are not prepared to deal with these risks because they still rely on traditional methods that cannot keep up with the speed of attacks led by artificial intelligence.”
The concern of governments and banks
The influence of Mythos is no longer limited to technical discussions among cybersecurity experts, as the US Secretary of the Treasury and the Chairman of the Federal Reserve have summoned a number of the largest US banks to discuss the risks that the model may pose to the financial system, according to the newspaper’s report.
The report also cited similar warnings from British and German officials, including an official at the British Ministry of Artificial Intelligence, Kanishka Narayan, who told the Financial Times that “we should be concerned” about the capabilities of this model.
As for the President of the German Central Bank, Joachim Nagel, he proposed making the “Methos” big data model available to the relevant institutions to ensure equal opportunities in this sector.
Chinese reports criticized the confinement of the Mythos model to a “closed club” of American companies, but local security companies saw this as an opportunity that could be used to develop national alternatives.
At the same time, James Gong, legal director at Bird & Bird law firm, believes that this technology may raise cybersecurity costs as companies increase their spending on employees, infrastructure and advanced protection systems, while creating new opportunities for security services based on artificial intelligence, according to what was reported by the South China Morning Post newspaper.
China between threat and opportunity
As for China, which is often described as one of the countries most exposed to cyber attacks, it cannot ignore these variables. Therefore, Chinese reports criticized the confinement of the Mythos model to a “closed club” of American companies, but local security companies saw this as an opportunity that could be invested in developing national alternatives.
According to Beijing-based consulting firm Concordia AI, Chinese open-source models still mostly lag behind closed-source American models in their cyber capabilities. However, the progress Chinese models have made over the past year has led to a rapid improvement in their capabilities as well.
The Global Times quoted a report by Chinese cybersecurity research group Natto Thoughts as saying that Chinese cybersecurity company 360 claimed to have developed an AI-powered “vulnerability discovery agent” that identified hundreds of previously unknown flaws, including in widely used software such as Microsoft Office and various types of Internet of Things devices.
Regarding vulnerability detection techniques, Guo Hongyi, founder of 360 Group, told a Global Times reporter that there are no predictable patterns in discovering system vulnerabilities using traditional methods, because it is essentially a matter of luck.
Guo warned that attackers with capabilities similar to Mythos’ capabilities will discover hundreds of times more vulnerabilities than expected when examining the target party’s shared software databases, which means that from the attacker’s perspective, the defender becomes like a filter full of holes (vulnerabilities).
Mythos may open the door to a “new cyber arms race” between the United States and China, as each side seeks to develop models capable of discovering and exploiting vulnerabilities on a large scale.
Some Chinese experts have described these smart models as “similar to a cyber-nuclear weapon,” and in the age of artificial intelligence, the weapon must be possessed so that the balance of this space is not broken.
Artificial Intelligence+
Recent developments have prompted some official authorities in China to take the model seriously, as Beijing commented on the “Methos” model directly for the first time in an article on the “Yueyuan Tantian” social media account run by the state-run Chinese Broadcasting Corporation, in which it stated that the model showed “unprecedented capabilities in cyberattacks.”
At the level of political discourse, Beijing has tried to integrate this debate into a broader vision of what it calls “artificial intelligence+,” as the People’s Daily, the mouthpiece of the Chinese Communist Party, published an analytical article entitled “How does artificial intelligence+ drive win-win global cooperation?”
The article emphasized that artificial intelligence “is no longer a separate technology, but has become a multiplier factor in all sectors, from industry to finance and health,” and that maximizing its benefits requires “common rules and harmonized standards in the areas of security, data, and ethics.”
The article stated that China calls for “global governance of artificial intelligence” based on “openness and win-win cooperation,” while emphasizing that military or sabotage uses of artificial intelligence in cyberspace pose a common danger to all countries, and no party can protect itself alone.
Arms race or opportunity for a new organization?
A series of South China Morning Post articles warned that Mythos could open the door to a “new cyber arms race” between the United States and China, as each side seeks to develop models capable of detecting and exploiting vulnerabilities on a large scale.
At a time when vulnerability detection models are arousing concern in Western capitals and stimulating the tightening of restrictions on sharing sensitive technologies, China finds its opportunity to present a relatively different narrative to transform the threat into an opportunity to build and manufacture advanced cybersecurity, by promoting the concept of artificial intelligence + as a framework for international cooperation in governance and setting standards.
Thus, the world faces a question that goes beyond the traditional debate about “the danger of artificial intelligence to jobs.” Does artificial intelligence remain just a tool in the hands of countries and companies, or have advanced vulnerability detection tools like Mythos begun to reshape the rules of the game in security, economics, and even politics?!
Between these two paths, it seems that the way the international community chooses to deal with artificial intelligence vulnerabilities today will determine the shape of cyberspace for decades to come, either as an arena for a digital arms race, or as a space of cooperation controlled by common rules.
