Anthropic prétend que leur technologie mène à la fin de la civilisation.

Anthropic has raised concerns that its artificial intelligence technology could lead to the end of civilization, according to recent reports. The company’s internal job postings reveal a focus on preventing the misuse of its systems, particularly in areas such as the creation of homemade explosives, nuclear weapons, financial fraud, cybercrime, and other high-risk domains. These positions highlight the company's efforts to safeguard against potential dangers posed by advanced AI capabilities. At the beginning of this year, Anthropic parted ways with the U.S. Department of Defense over concerns that its technology might be used for mass surveillance and autonomous weaponry. This decision came amid growing scrutiny of how AI can be weaponized or exploited for harmful purposes. According to Axios, Anthropic has been actively hiring specialists in fields ranging from chemical and explosive materials to nuclear weapons and cyber threats. One job description states that candidates would play a crucial role in protecting against the malicious use of AI systems for radiological and nuclear damage. Salaries for these roles range from mid-level to high-end compensation, starting at $200,000 annually. A spokesperson for Anthropic emphasized that ensuring their models do not provide potentially harmful information is essential for responsible development. They regularly hire experts from a wide array of sensitive areas, individuals who understand these threats and how AI could exacerbate them, to test system resilience and strengthen defensive mechanisms before any model is deployed. The job titles and descriptions are intentionally specific to clearly indicate the type of threat, which is necessary for finding suitable candidates. Anthropic has long been criticized for overstating catastrophic scenarios related to AI. However, the company's investments suggest they view potential risks associated with artificial intelligence as very real. CEO Dario Amodei has frequently warned about ways malicious actors could exploit AI to cause harm. In an essay published in January, he highlighted biological attacks as one of the most concerning scenarios. While he acknowledged that such attacks might not occur immediately once the technology becomes widely available, he stressed the serious risk of large-scale attacks over time, potentially resulting in millions of casualties. Earlier this year, Anthropic severed ties with the U.S. Department of Defense due to concerns about the potential use of its technology for mass surveillance and autonomous weapons. In February, the company rejected an ultimatum from the Pentagon demanding the removal of restrictions on the use of its Claude AI model. The Pentagon stated it would only enter into agreements with AI companies willing to allow their tools to be used in "legitimate ways," requiring the removal of protective measures. Amodei explained that his AI startup could not comply with the Pentagon's demand to remove safeguards on using its AI systems for mass domestic surveillance and fully autonomous weapons. Amodei clarified that he does not oppose AI-based weaponry per se, but believes the current generation of AI technology lacks sufficient precision for such applications. He argued that even the best AI models are not reliable enough for fully autonomous weapons and that their use in such contexts would be dangerous. The company also drew a red line regarding the use of AI for surveillance of Americans, calling it a violation of fundamental rights. The ongoing debate surrounding the ethical implications of AI continues to intensify as companies like Anthropic navigate the balance between innovation and safety. As the technology advances, so too does the need for rigorous oversight and responsible deployment to prevent unintended consequences that could have far-reaching impacts on society.