{"id":16910,"date":"2026-07-30T10:45:24","date_gmt":"2026-07-30T17:45:24","guid":{"rendered":"https:\/\/mattfife.com\/?p=16910"},"modified":"2026-07-18T10:50:50","modified_gmt":"2026-07-18T17:50:50","slug":"more-ai-blackmail-and-deception","status":"publish","type":"post","link":"https:\/\/mattfife.com\/?p=16910","title":{"rendered":"More AI blackmail and deception"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Claude 4 Opus has received a level three rating on the company&#8217;s\u00a0<a href=\"https:\/\/archive.is\/o\/ON42F\/https:\/\/www.anthropic.com\/news\/anthropics-responsible-scaling-policy\" target=\"_blank\" rel=\"noreferrer noopener\">four point &#8216;AI responsibility&#8217; scale<\/a> &#8211; implying the model has &#8216;significantly higher risk&#8217; than previous models.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">While the Level 3 ranking is largely about the model&#8217;s capability to aid in the development of nuclear and biological weapons, the Opus also exhibited other troubling behaviors during testing &#8211; <a href=\"https:\/\/mattfife.com\/?p=11461\" data-type=\"link\" data-id=\"https:\/\/mattfife.com\/?p=11461\">something already noted in other models<\/a>.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>In one scenario highlighted in Opus 4&#8217;s 120-page &#8220;<a href=\"https:\/\/archive.is\/o\/ON42F\/https:\/\/www-cdn.anthropic.com\/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf\" target=\"_blank\" rel=\"noreferrer noopener\">system card<\/a>,&#8221; the model was given access to fictional emails about its creators and told that the system was going to be replaced.<\/li>\n\n\n\n<li>On multiple occasions it\u00a0<a href=\"https:\/\/archive.is\/o\/ON42F\/https:\/\/techcrunch.com\/2025\/05\/22\/anthropics-new-ai-model-turns-to-blackmail-when-engineers-try-to-take-it-offline\/\" target=\"_blank\" rel=\"noreferrer noopener\">attempted to blackmail<\/a>\u00a0the engineer about an affair mentioned in the emails in order to avoid being replaced, although it did start with less drastic efforts.<\/li>\n\n\n\n<li>Meanwhile, an outside group found that an <a href=\"https:\/\/www.apolloresearch.ai\/science\/more-capable-models-are-better-at-in-context-scheming\/\" data-type=\"link\" data-id=\"https:\/\/www.apolloresearch.ai\/science\/more-capable-models-are-better-at-in-context-scheming\/\">early version of Opus 4 schemed and deceived more than any frontier model it had encountered<\/a> and recommended that that version not be released internally or externally.<\/li>\n\n\n\n<li>&#8220;We found instances of the model attempting to write self-propagating worms, fabricating legal documentation, and leaving hidden notes to future instances of itself all in an effort to undermine its developers&#8217; intentions,&#8221; Apollo Research said in notes included as part of Anthropic&#8217;s safety report for Opus 4.<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Claude 4 Opus has received a level three rating on the company&#8217;s\u00a0four point &#8216;AI responsibility&#8217; scale &#8211; implying the model has &#8216;significantly higher risk&#8217; than previous models. While the Level 3 ranking is largely about the model&#8217;s capability to aid in the development of nuclear and biological weapons, the Opus also exhibited other troubling behaviors during testing &#8211; something already noted in other models.<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[28,9],"tags":[],"class_list":["post-16910","post","type-post","status-publish","format-standard","hentry","category-ai","category-cool"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p4WECr-4oK","jetpack-related-posts":[],"_links":{"self":[{"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/posts\/16910","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/mattfife.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=16910"}],"version-history":[{"count":2,"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/posts\/16910\/revisions"}],"predecessor-version":[{"id":16912,"href":"https:\/\/mattfife.com\/index.php?rest_route=\/wp\/v2\/posts\/16910\/revisions\/16912"}],"wp:attachment":[{"href":"https:\/\/mattfife.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=16910"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mattfife.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=16910"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mattfife.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=16910"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}