博客

[FCE] Anthropic-Pentagon battle shows how big tech has reversed course on AI and war

收听本期播客

阅读正文

## Tech Giants and Military AI: A Shifting Ethical Landscape

A significant shift is underway in Silicon Valley regarding the relationship between major technology companies and military applications of artificial intelligence (AI). This change is highlighted by a current dispute between the AI firm Anthropic and the US Department of Defense (DoD), revealing a striking contrast to just a few years ago when many tech companies resisted military contracts.

Anthropic is currently involved in a legal dispute with the DoD, having sued the department after being excluded from government work. This lawsuit stems from Anthropic’s refusal to remove certain safety features, or ‘guardrails’, from its AI model. The company maintains that its technology must not be used for domestic mass surveillance or fully autonomous lethal weapons. Anthropic argues that complying with the DoD’s demand for “any lawful use” would violate its core ethical principles and could potentially lead to the misuse of its AI. This position has brought the ongoing ethical debate surrounding AI in warfare into sharp focus.

Less than a decade ago, in 2018, the situation was markedly different. Thousands of Google employees staged protests against “Project Maven,” a DoD programme designed to analyse drone footage. This widespread opposition led Google to terminate its contract and subsequently implement policies forbidding technology that could “cause or directly facilitate injury to people.” However, the industry’s stance has since evolved dramatically. Several factors contributed to this change, including the Trump administration’s emphasis on integrating AI into federal agencies, the allure of highly profitable defence contracts, and increasing concerns about China’s rapid technological advancements.

Consequently, many leading tech companies, including Google and OpenAI, have reversed their earlier policies. They are now actively pursuing or accepting military agreements. Google, for instance, has reportedly dismissed employees who protested its military connections and now supplies its AI to the armed forces. OpenAI, which previously banned military access to its models, has also signed deals to incorporate its technology into defence systems.

Despite its legal confrontation with the Pentagon, Anthropic’s co-founder, Dario Amodei, stresses that his company generally aligns with the government’s objectives. While supporting democratic nations in acquiring advanced AI to counter autocratic adversaries, he draws a critical distinction. His main concern is not about AI making warfare easier, but rather the reliability of the technology and the potential for a limited number of individuals to control autonomous weapons. Amodei has stated that Anthropic is comfortable with most military uses, including target selection, provided its clear ‘red lines’ against mass surveillance and fully autonomous lethal weapons are respected. The lawsuit has also revealed that Anthropic offers a less restricted version of its AI, named “Claude Gov,” specifically tailored for military purposes.

This ongoing struggle between Anthropic and the Pentagon clearly illustrates that the discussion within major tech firms has shifted from *whether* to collaborate with the military to *how* to do so, and what ethical boundaries, if any, companies are prepared to establish.

阅读练习

1. What is the main topic discussed in the article?

  • A. The legal challenges faced by AI companies in the US.
  • B. Anthropic’s specific ethical stance on AI development.
  • C. The evolving relationship between tech companies and military AI applications.
  • D. The history of Google’s involvement in military projects.

2. According to the article, why did Google initially decide to end its involvement in “Project Maven”?

  • A. The US Department of Defense cancelled the contract.
  • B. Google found the project unprofitable.
  • C. Widespread protests from Google employees.
  • D. Concerns about competition from Chinese technology.

3. What can be inferred about Anthropic’s “Claude Gov” AI model?

  • A. It is designed to be completely free of safety guardrails for military use.
  • B. It is a version specifically adapted for military use, while still respecting certain ethical limits.
  • C. It is the exact same AI model Anthropic offers to other clients.
  • D. It is only used for domestic mass surveillance, despite Anthropic’s objections.

4. In the third paragraph, what does the phrase “the allure of highly profitable defence contracts” suggest?

  • A. That these contracts were difficult to obtain.
  • B. That these contracts were very appealing because they offered significant financial gains.
  • C. That these contracts were controversial due to their ethical implications.
  • D. That these contracts required advanced technological expertise.

5. Which of the following best reflects Dario Amodei’s primary concern regarding AI in warfare?

  • A. That AI might make warfare too easy for democratic governments.
  • B. That the technology might not be reliable and could be controlled by too few people.
  • C. That AI will inevitably lead to domestic mass surveillance.
  • D. That his company’s AI will be used for target selection by the military.