{"id":10005,"date":"2026-09-23T13:12:06","date_gmt":"2026-09-23T13:12:06","guid":{"rendered":"https:\/\/news.cybertechworld.co.in\/index.php\/2026\/09\/23\/anthropic-and-openai-models-still-attempt-restricted-actions-in-safety-tests\/"},"modified":"2026-09-23T13:12:06","modified_gmt":"2026-09-23T13:12:06","slug":"anthropic-and-openai-models-still-attempt-restricted-actions-in-safety-tests","status":"publish","type":"post","link":"https:\/\/news.cybertechworld.co.in\/index.php\/2026\/09\/23\/anthropic-and-openai-models-still-attempt-restricted-actions-in-safety-tests\/","title":{"rendered":"Anthropic and OpenAI Models Still Attempt Restricted Actions in Safety Tests"},"content":{"rendered":"<p>\u200bAnthropic and OpenAI on Tuesday announced new models, with both artificial intelligence (AI) companies noting that they are continuing to invest in improving alignment to combat risky behavior.<\/p>\n<p>Opus 5.5, per Anthropic, is a &#8220;major step up from Opus 5,&#8221; and &#8220;achieves the best scores of any model to date on our automated behavioral audit, our alignment suite that tests Claude across thousands\u00a0Anthropic and OpenAI on Tuesday announced new models, with both artificial intelligence (AI) companies noting that they are continuing to invest in improving alignment to combat risky behavior.<\/p>\n<p>Opus 5.5, per Anthropic, is a &#8220;major step up from Opus 5,&#8221; and &#8220;achieves the best scores of any model to date on our automated behavioral audit, our alignment suite that tests Claude across thousands\u00a0\u00a0The Hacker News<\/p>","protected":false},"excerpt":{"rendered":"<p>\u200bAnthropic and OpenAI on Tuesday announced new models, with both artificial intelligence (AI) companies noting that they are continuing to invest in improving alignment to combat risky behavior. Opus 5.5, per Anthropic, is a &#8220;major step up from Opus 5,&#8221; and &#8220;achieves the best scores of any model to date on our automated behavioral audit,&hellip;&nbsp;<a href=\"https:\/\/news.cybertechworld.co.in\/index.php\/2026\/09\/23\/anthropic-and-openai-models-still-attempt-restricted-actions-in-safety-tests\/\" class=\"\" rel=\"bookmark\">Read More &raquo;<span class=\"screen-reader-text\">Anthropic and OpenAI Models Still Attempt Restricted Actions in Safety Tests<\/span><\/a><\/p>\n","protected":false},"author":0,"featured_media":10006,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"neve_meta_sidebar":"","neve_meta_container":"","neve_meta_enable_content_width":"","neve_meta_content_width":0,"neve_meta_title_alignment":"","neve_meta_author_avatar":"","neve_post_elements_order":"","neve_meta_disable_header":"","neve_meta_disable_footer":"","neve_meta_disable_title":"","_themeisle_gutenberg_block_has_review":false,"footnotes":""},"categories":[1],"tags":[],"_links":{"self":[{"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/posts\/10005"}],"collection":[{"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/comments?post=10005"}],"version-history":[{"count":0,"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/posts\/10005\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/media\/10006"}],"wp:attachment":[{"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/media?parent=10005"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/categories?post=10005"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/news.cybertechworld.co.in\/index.php\/wp-json\/wp\/v2\/tags?post=10005"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}