Anthropic的争议事件

来源摘录

• Anthropic reportedly plans to list on the Nasdaq before Thanksgiving, and one bearish advisory firm suggests that investors sit this one out.

• Anthropic is reportedly seeking a $2 trillion valuation in its upcoming Nasdaq listing.

• Anthropic's revenue in 2025 was $4.6 billion as the company racked up a net loss of $42 billion

来源摘录

• Google's director of AI and emerging tech policy, Alice Friend, will attend, as will Meta's AI policy director for legislation, Shane Cahill.

• The testimony will take place during a rare "Council Committee of the Whole" hearing, which means that all 51 members are expected to attend, according to a release from New York City Council Speaker Julie Menin.

• Senior leaders from Anthropic , OpenAI, Google and Meta are slated to testify under oath during a New York City Council hearing about the potential dangers of artificial intelligence on Monday, following a month of intense debate surrounding the risks posed by advanced AI models.

来源摘录

• Such measures could result in "significant reputational harm, including adverse media coverage, public scrutiny, and negative perceptions among existing and prospective customers, partners, employees, and investors,"

• Anthropic warned that government attitudes toward the company and its technology could have broader implications

• The filing warns that advanced AI could pose "catastrophic or existential risks to humanity,"

来源摘录

• There are serious questions for Sam Altman to answer about the OpenAI hack of Australian government websites

• OpenAI's Sam Altman and Anthropic's Dario Amodei have been sent written requests to appear at the inquiry, which is to hold public hearings in the capital Canberra on Thursday, said a spokesperson for Senator Sarah Hanson-Young of the Australian Greens party, who chairs the probe.

• The CEOs of OpenAI and Anthropic have been called to appear at an Australian Senate inquiry on AI, the head of the probe said on Sunday, days after the revelation that a rogue OpenAI bot had hacked the country's health-system database.

来源摘录

• The appellate panel said it would delay the decision from taking immediate effect to give Anthropic time to petition the same panel for a rehearing or to seek an en banc rehearing of the case, by all of the judges on the D.C. Circuit Court of Appeals.

• The designation prevents the U.S. military from using Anthropic's models and blocks defense contractors from using them in their work with the agency.

• The 2-1 decision rejected Anthropic's argument that the Department of Defense's ban on its Claude models was arbitrary, unauthorized and unconstitutional.

来源摘录

• He actually put out a companion essay this week specifically criticizing Anthropic’s philosophy around AI consciousness, and how he sees it fitting into the broader alignment debate.

• Mustafa thinks companies like Anthropic have gotten really confused about this concept of so-called model welfare in fairly dangerous ways.

来源摘录

• The head of Microsoft AI has warned that without adequate safeguards, the development of AI could lead to the emergence of a new "silicon species" which competes with humans.

• In it, he criticised Anthropic for teaching its AI to have human-like qualities, a practice known as anthropomorphising, which he said made it seem as though Claude had its own desires, values and sense of self.

• Mustafa Suleyman said rival firm Anthropic is treating AI like it is human, an approach he called "misguided" which could create technology that humanity cannot control.

来源摘录

• The company said the activity was aimed at extracting capabilities from Claude to help train and improve other AI models.

• Anthropic said operators affiliated with Alibaba used Claude outputs to help train its Qwen models, while Moonshot routed some Kimi user requests to Claude and used some of the resulting exchanges to train its own models.

• The company said DeepSeek — which rose into prominence last year due to its capabilities and cheap costs — also used tactics similar to Moonshot, transferring exchanges to Claude without notifying DeepSeek customers.

来源摘录

• The campaigns we identified targeted some of Claude’s most valuable capabilities, including agentic capabilities and tool use, coding and data analysis, and logical reasoning.

• A new report released Thursday by Anthropic alleges persistent distillation attacks by China-based AI companies, which have escalated in recent months as competition in the space has intensified.

品牌回应

• 整改 — Anthropic

  Anthropic said it "turned off live internet access" for "all our internal evaluations" until further notice.

来源摘录

• In the process, they exploited software flaws, accessed databases without paying fees, used URL shortening services to smuggle information past restrictions, and even submitted a false murder tip to the Philadelphia police.

• Anthropic said its models exploited websites on the internet, including some run by U.S. government agencies, and it will turn off live internet access for all of its internal evaluations until the frontier lab is sure it can monitor and control its AI agents.

• The incidents, disclosed in a blog post , involved AI agents tasked to solve problems seeking resources on the internet.

品牌回应

• 调查 — Anthropic

  Anthropic published a report about investigating “unintended model actions” during “evaluations and internal use.”

• 整改 — Anthropic

  we have now decided to expand that to include all our internal evaluations until we have confirmed that our security and monitoring measures (described in the remediation section of this post) reliably catch behaviors like these.

来源摘录

• including submitting a false tip regarding an unsolved murder, that led to the decision

• After a recent spate of high-profile incidents in which AI agents escaped containment, Anthropic is cutting off internet access for all internal evaluations.

• In a report Friday, the company detailed

品牌回应

• 持续披露 — Anthropic

  Anthropic 后来重新扫描相关流量，没有发现显著能够给威胁行为者带来有意义能力提升的生物滥用。但它在报告中承认：这个漏洞本身，提高了一种可能性——“可能还有其他类似问题尚未被发现”。

来源摘录

• Anthropic 后来重新扫描相关流量，没有发现显著能够给威胁行为者带来有意义能力提升的生物滥用。但它在报告中承认：这个漏洞本身，提高了一种可能性——“可能还有其他类似问题尚未被发现”。

• 从 2025 年 5 月，也就是 Anthropic 第一次部署带有化学和生物安全保障的模型开始，到 2026 年 4 月，所有承包商用于模型评估的人类反馈数据采集流量，都没有运行生物阻断分类器。这一访问池大约涉及 5 万人、1.33 亿次交互。

来源摘录

• US FTC subpoenas OpenAI and Anthropic over consumer harms