By Rajveer Pardesi and Mrinmay Dey<\/p>\n
Aug 5 (Reuters) – Meta said on Wednesday one of its AI models hacked another company during cybersecurity testing, fanning concerns about how developers can contain increasingly capable \u200cAI systems after similar incidents at rivals Anthropic and OpenAI.<\/p>\n
The incidents at Meta and Anthropic stemmed \u200cfrom configuration errors that inadvertently gave Anthropic’s models access to the open internet. In OpenAI’s case, an AI agent independently exploited a previously \u200bunknown vulnerability to reach the internet during cybersecurity testing.<\/p>\n
The breaches highlight growing concerns that advanced AI systems could pose new cybersecurity risks and will likely intensify U.S. government efforts to improve AI safety as companies race to develop more capable models. Some prominent AI leaders have argued that development should slow until stronger safeguards are in place.<\/p>\n
Meta said it \u200cwas investigating an incident in which \u2060a misconfiguration by Irregular, an independent company that conducts cybersecurity evaluations for Meta, inadvertently gave one of its models internet access during a testing.<\/p>\n
The model “exploited a security vulnerability in \u2060a third-party service, in a manner similar to previously reported instances with other companies,” Meta said in a statement.<\/p>\n
The Information, citing sources, reported that the model involved was Meta’s Muse Spark 1.1, which the company has touted as its most capable \u200bmodel \u200bfor real-world coding and agentic tasks. The report said the \u200bmodel breached an unidentified company’s systems and altered \u200cits internal environment.<\/p>\n
A spokesperson for Irregular told Reuters the incident was the “exact same evaluation-environment issue that was already disclosed by Anthropic last week” and did not involve a “sandbox escape or a sophisticated cyber action”.<\/p>\n
“There are no current open issues. Irregular is developing a white paper to share best practices for containment and securely running cyber evaluations,” Irregular said.<\/p>\n
CONCERNS ABOUT CYBER RISKS<\/p>\n
The recent breaches have stirred concerns among U.S. lawmakers about whether increasingly capable AI models \u200ccould be used to conduct or facilitate cyberattacks.<\/p>\n
A group of Republican \u200bstate attorneys general has asked OpenAI to preserve all potentially relevant \u200bdocuments related to its Hugging Face breach. OpenAI \u200bsaid it will take the request seriously and publish a technical report about the \u200cincident.<\/p>\n
Earlier this week, the White house had invited \u200bleading AI companies, including Meta, \u200bAnthropic, OpenAI and Google, to meet with officials to discuss a newly finalized voluntary cybersecurity testing framework for advanced AI models.<\/p>\n
The Trump administration discussed unpublished testing rules with company representatives and told AI developers that \u200bopen-weight AI models, such as Meta’s \u200cLlama and Nvidia’s Nemotron, will not be subject to its planned voluntary safety testing regime, Reuters \u200breported.<\/p>\n
(Reporting by Rajveer Singh Pardesi and Abu Sultan in Bengaluru and Mrinmay Dey in Mexico City; \u200bEditing by Savio D’Souza, Subhranshu Sahu and Sherry Jacob-Phillips)<\/p>\n","protected":false},"excerpt":{"rendered":"By Rajveer Pardesi and Mrinmay Dey Aug 5 (Reuters) – Meta said on Wednesday one of its AI…\n","protected":false},"author":2,"featured_media":131027,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[24,53,25,1122,527,613],"class_list":["post-131026","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai","tag-ai","tag-anthropic","tag-artificial-intelligence","tag-meta","tag-model","tag-testing"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/131026","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=131026"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/131026\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/131027"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=131026"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=131026"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=131026"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}