"Wake up and the world has changed"! Mythos AI model triggers panic among global financial executives, Anthropic shifts focus to "weakened version" Opus 4.7 to explore commercialization.

The Middle East conflict and the historic energy impact have been major risks threatening financial stability in recent times and cannot be ignored. However, the yet-to-be-released artificial intelligence (AI) model Mythos by Anthropic has added another layer of concern for policymakers at two international conferences in Washington. These conferences were hosted by two major institutions established eighty years ago with the aim of promoting global economic harmony. At the International Monetary Fund and World Bank Spring Meetings, the focus of discussion shifted to the Middle East military conflict, prompting governments to restart crisis response plans, commit to cooperation within their capabilities, and boost consumption through fiscal and monetary policy tools. However, there were more questions than answers surrounding Mythos. Christine Lagarde, the President of the European Central Bank, expressed in an interview on Tuesday, "If this technology falls into the hands of criminals, the consequences could be very severe." She was referring to the potential destructiveness of Mythos. This concern has been echoed by central bank governors, finance ministers, regulatory agencies, and investment executives, who have been exploring the potential risks and protective measures of this technology. The concerns surrounding this issue came to the public's attention earlier this month when US Treasury Secretary Benson summoned Wall Street executives, warning them of an upcoming AI model that could usher in a new era of network attacks launched autonomously by Siasun Robot & Automation. Benson stated in an interview on Wednesday, "I am confident that all parties have reached a consensus to work together to enhance the resilience of the financial system." However, outside the United States, the sense of urgency conveyed at these conferences has put senior officials and bankers on high alert. Some worry that this technology might breach traditional network defense systems, exposing the financial system to incalculable threats. Andrew Bailey, Governor of the Bank of England, remarked earlier this week at an event at Columbia University in New York, "In theory, the events in the Gulf region are the most recent challenges facing the world today." But, he continued, "you may wake up to find that Anthropic has found a way to completely disrupt the cybersecurity landscape, and you can't help but wonder what you did wrong in your past life." Core Questions Bailey and other experts in financial stability face the challenge of knowing little about the threats posed by Mythos, unable to determine the extent to which this model has surpassed the network security risks that have been on their radar for years. Will Mythos trigger large-scale bank account fund thefts, international payment system paralysis, or precipitate a crisis of trust, shaking the foundation of the financial system - credit? Is this risk similar to the known threats posed by quantum computing (which regulatory agencies have extensively researched), or does it require entirely new countermeasures? These questions linger in the corridors and meeting rooms of the Willard Hotel in Washington. This hotel, where Abraham Lincoln resided before his inauguration, hosted a conference for senior officials and executives organized by the International Finance Association this week. Many questions remain unanswered to this day. It is unclear how much detail and access Anthropic, based in San Francisco, has shared with non-US officials and banks regarding the Mythos technology. A regional head of a European bank stated that they have been promised to attend a briefing on the situation. Limited channels for information gathering may lead to difficulties for some parts of the global financial system to follow the recommendations of the International Monetary Fund, always staying "close to the cutting edge of cyber threats." Carson Cheng, Finance Minister of Canada, stated in an interview on Wednesday that Mythos "requires our full attention," as he hopes to discuss this issue with finance ministers from around the world. He said, "Safeguarding the resilience of the financial system is in our common interest." An insider revealed this week that the US Treasury's technology team is working to gain access to Anthropic's Mythos AI model for vulnerability assessment. John Williams, President of the Federal Reserve Bank of New York, stated on Thursday that recent advances in AI technology "have shown the powerful capabilities of such tools, both in identifying vulnerabilities and exploiting them, far beyond many people's expectations." The New York Fed is a central hub for global financial transactions and stability risk regulation. Role of Financial Stability Committees Traditionally, financial stability-related risks have been coordinated by financial stability committees, with the current chairman being Andrew Bailey, Governor of the Bank of England. However, in the field of cyber risks, financial stability committees are often in a secondary position, as some member countries are reluctant to disclose relevant information within this broad alliance. This leads to cyber security issues predominantly being led by smaller organizations such as the Group of Seven. According to two insiders, during the Group of Seven finance ministers meeting held on Wednesday, participants discussed the importance of establishing an international regulatory framework and standardizing AI governance, with the consensus being that no country can address the risks posed by Mythos alone. One source added that given the exponentially growing risks and potential severe consequences, all parties support delving deeper into this issue, although the specific path for progress is not yet clear. Even more concerning is that this doomsday-level AI technology, coupled with heightened geopolitical risks, may transform from a tool to enhance efficiency in the financial industry to a blade that shreds its technological infrastructure. The Middle East conflict exacerbates countries' concerns about state-sponsored cyber attacks. "Urgency is paramount" Officials and bankers alike are acutely aware that if US companies can develop such destructive AI technology, then equally hostile countries may achieve breakthroughs without alerting Wall Street in advance. Laila Khawaja, Managing Director of Gavekal Technologies, who specializes in researching US-China tech issues, stated, "The speed of development of cutting-edge AI technology has far exceeded the pace of regulatory framework implementation. Before these models deeply integrate into critical sectors like finance, building defenses is of utmost importance." This week's quest for answers also highlights the growing "wealth gap" in the AI domain and the fragmentation of the global financial stability system. Europe is increasingly concerned about its over-reliance on overseas tech companies. Officials privately lament that the prevailing "every man for himself" mentality in global financial discussions may hinder the effective sharing of global risk information. Several European officials expressed their intent to urge the US to share relevant information through this week's meetings. According to Elizabeth Swantson, Finance Minister of Sweden, a warning meeting involving central bank governors and finance ministers will be held later on Thursday, with AI being one of the main agenda items. She stated, "The point I want to convey is that AI itself is extremely valuable, but the uncontrollable dominant force overlaid with AI is dangerous." She added that the current core issues are the US-Israeli conflict, uncertainty in the situation, and cybersecurity threats. Anthropic promotes a "weakened version" AI model to pave the way for commercialization Just as the global financial industry debates the safety boundaries of Mythos, Anthropic officially launched an upgraded version of its flagship model - Opus 4.7 - on Thursday. The company claims that Opus 4.7 performs better in the field of software engineering, handling high-difficulty programming tasks that previously required more human supervision, with improved obedience to commands and image recognition accuracy. Of particular note, Anthropic explicitly stated in its announcement that Opus 4.7 has a weaker overall capability in network security compared to Mythos, and the company even experimentally reduced its network attack and defense capabilities during the training phase. Additionally, the model has built-in automatic detection and interception systems to block requests involving prohibited or high-risk cybersecurity uses. This deployment strategy reflects Anthropic's cautious balance between cutting-edge capabilities and security boundaries. The company's statement mentioned that the experience accumulated from the actual deployment of Opus 4.7 will pave the way for the widespread commercialization of a Mythos-level model. Currently, Mythos is only open to select corporate clients to assist them in fortifying their systems. This move also aligns with market expectations for the progress of Anthropic's commercialization efforts. In the midst of intense competition with OpenAI, Anthropic is seeking a new round of funding, with its valuation expected to double from the previous round's $380 billion to $800 billion or even higher. Its AI programming products are gaining momentum, and consumer interest is steadily increasing.