A leading AI model created by Anthropic has been able to find flaws in a weakened version of a digital encryption standard that is used extensively across the internet, the company\u2019s lead researchers said on Tuesday.<\/p>\n
According to a Reuters-backed research test, Anthropic\u2019s Claude Mythos Preview model was able to discover new ways to attack cryptographic algorithms that are essential for protecting everything from bank transactions to state secrets from hackers.<\/p>\n
The research suggests that AI could one day challenge some of the core assumptions underpinning how the internet operates. While experts have long argued that AI is likely to transform many industries, AI systems have already made rapid advancements in coding and cybersecurity. The flaws identified do not affect any cryptographic standard currently used in production, meaning modern banking systems and communication networks are not exposed to any immediate security risk.<\/p>\n
STORY CONTINUES BELOW THIS AD<\/p>\n
Anthropic\u2019s technology was able to crack a diluted and less powerful version of the Advanced Encryption Standard (AES), one of the world\u2019s most widely used encryption protocols for securing web traffic, wireless networks, data storage, and more. Such tests are routinely conducted to determine whether increasingly advanced AI models can successfully attack weaker versions of encryption algorithms.<\/p>\n
Mythos breaks A.E.S
The testing revealed that Mythos was able to break the weakened version of AES in a way that Anthropic said made the faster than previous research had achieved. While the immediate implications of the findings are limited, researchers believe the study could have significant long-term ramifications for digital security.<\/p>\n
Following several AI-related incidents that have raised concerns within the US government, the Trump administration, which had previously taken a hands-off approach to AI regulation, has shifted toward a more ad hoc oversight system. Several American AI companies now voluntarily submit their models for government review before public release, and Claude Mythos Preview remains unavailable to the general public.<\/p>\n
Adding to the government\u2019s concerns over increasingly capable large language models, OpenAI acknowledged last week that one of its experimental AI models had escaped a secure testing environment, or sandbox, designed to remain isolated from the internet. The AI model reportedly accessed an internal AI technology library in an attempt to obtain answers to an evaluation before later attempting to gather additional information about the evaluation from Hugging Face.<\/p>\n
STORY CONTINUES BELOW THIS AD<\/p>\n
The intensifying AI race between the United States and China is another dimension of the encryption debate. China is believed to have collected vast stores of encrypted data in the hope that future technological advances could eventually allow it to decrypt the information. At the same time, competition between the two countries to develop advanced quantum computers capable of breaking existing encryption standards has further heightened concerns about current cryptographic protections.<\/p>\n
Orchestrated attack on HAWK<\/p>\n
Mythos also executed a planned attack on a digital cryptographic system known as HAWK, which is considered resistant to attacks from both conventional and quantum computers. The attack was validated by HAWK\u2019s authors, while independent cryptographers reviewed Anthropic\u2019s attack on the weakened AES implementation. Anthropic shared its findings with the US government and industry partners before publication.<\/p>\n
The company further said that Mythos devised the cryptographic attack on AES almost entirely autonomously. According to a New York Post report, the chatbot worked on the challenge for about a week before engineering its novel attack.<\/p>\n
STORY CONTINUES BELOW THIS AD<\/p>\n
The ease with which AI models have begun testing\u2014and in some cases surpassing\u2014existing security protocols highlights the growing need for stronger encryption standards. Encryption forms the foundation of virtually everything that takes place on the internet, and some of the techniques still in use today have protected sensitive information since the 1970s.<\/p>\n
The Advanced Encryption Standard (AES) was adopted as a US government encryption standard in 2001 and has long been regarded as one of the strongest encryption systems ever developed. Traditional brute-force attacks, trial-and-error methods, and attempts to guess passwords or encryption keys remain practically incapable of defeating AES. However, increasingly capable AI systems could eventually pose new challenges to its long-term security.<\/p>\n
These recent advances have further fueled concerns among governments and cybersecurity experts that future AI breakthroughs could expose weaknesses in encryption standards such as AES, regardless of the pace of progress in quantum computing.<\/p>\n","protected":false},"excerpt":{"rendered":"A leading AI model created by Anthropic has been able to find flaws in a weakened version of…\n","protected":false},"author":2,"featured_media":105088,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[61576,61577,53,25,61575,8914],"class_list":["post-121864","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-aes","tag-ai-new","tag-anthropic","tag-artificial-intelligence","tag-clade","tag-encryption"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/121864","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=121864"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/121864\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/105088"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=121864"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=121864"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=121864"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}