博客

[FCE] Goldman Sachs chief ‘hyper-aware’ of risks from Anthropic’s Mythos AI

收听本期播客

阅读正文

The financial sector is increasingly facing significant cybersecurity challenges as advanced artificial intelligence models develop at an unprecedented pace. David Solomon, the chief executive of Goldman Sachs, has expressed profound awareness of the capabilities of Anthropic’s new AI model, Mythos, revealing that the US banking giant is now collaborating closely with the tech firm following serious warnings about the model’s potential threats.

Major US banks, including Goldman Sachs, have been closely observing the rapid progress in AI, particularly large language models, as part of their broader strategy to safeguard against cyber-attacks. During a recent earnings call, Solomon informed analysts that these new AI models possess ‘enhanced capabilities’. He explained that Goldman Sachs is addressing these advancements through cooperation with both the US government and the developers of the models themselves.

Concerns primarily centre on Anthropic’s latest creation, Mythos. The company itself issued a blog post last week, claiming that Mythos presents an unprecedented risk to digital security. The post detailed that Mythos has achieved a level of coding proficiency that allows it to surpass most human experts in identifying and exploiting weaknesses within IT systems. The potential ramifications for national security, public safety, and global economies are described as extremely serious.

In response to these revelations, Solomon confirmed Goldman Sachs’ active engagement with Anthropic and all its security vendors to effectively manage these advanced capabilities. He emphasised the bank’s growing and crucial focus on strengthening its cyber and infrastructure resilience. This heightened attention comes after a meeting last week where US Treasury Secretary Scott Bessent convened Solomon and other leaders from ‘systemically important banks’ – institutions whose failure could jeopardise overall financial stability – to specifically discuss the Mythos model.

Across the Atlantic, the UK government’s AI Security Institute (AISI) has also issued a warning regarding Mythos, describing it as a significant ‘step up’ in cyber threat. AISI highlighted its capacity to autonomously execute multi-step attacks and uncover IT system vulnerabilities, tasks that typically demand days for human professionals. Tests conducted by AISI demonstrated Mythos successfully completing a complex 32-step cyber-attack simulation in three out of ten attempts.

AISI concluded its report by underlining that future advanced AI models are expected to further improve on Mythos’s capabilities, making immediate investment in cyber defence absolutely vital. UK regulators are likewise preparing to address these emerging risks with British bank executives and government officials in the coming weeks. Societies globally now face the complex task of balancing the immense potential of advanced AI with the critical necessity of protecting digital security and infrastructure.

阅读练习

1. What is the main purpose of this article?

  • A. To explain how advanced AI models are being used to protect banks from cyber-attacks.
  • B. To describe the growing cybersecurity risks posed by advanced AI and the responses from financial institutions and governments.
  • C. To promote Anthropic’s new AI model, Mythos, as a powerful technological breakthrough.
  • D. To compare the cybersecurity strategies of US and UK banks in detail.

2. According to the article, what is a key capability of Anthropic’s Mythos AI model?

  • A. It can predict future cyber-attacks with 100% accuracy.
  • B. It can identify and exploit weaknesses in IT systems more effectively than most human experts.
  • C. It is designed to assist banks in developing stronger cybersecurity defences.
  • D. It has successfully completed all 32 steps of a complex cyber-attack simulation.

3. The meeting organised by US Treasury Secretary Scott Bessent was held to discuss:

  • A. The overall financial stability of ‘systemically important banks’.
  • B. Goldman Sachs’s investment in new AI security vendors.
  • C. The potential threats posed by the Mythos AI model to the financial sector.
  • D. Plans for US banks to collaborate more closely with UK regulators on AI security.

4. What did the UK’s AI Security Institute (AISI) highlight about Mythos?

  • A. Its ability to create new, advanced AI models independently.
  • B. Its capacity for autonomous multi-step cyber-attacks and vulnerability discovery.
  • C. Its success in completely preventing all cyber threats in test environments.
  • D. That it requires extensive human supervision to perform complex tasks.

5. The article mentions ‘systemically important banks’ are those:

  • A. That have significantly invested in advanced AI technology.
  • B. Whose operations are critical to the broader financial health and stability.
  • C. That are under direct threat from the Mythos AI model.
  • D. That primarily serve government agencies and national security interests.