{"version":"1.0","type":"rich","provider_name":"techandbusiness.org","provider_url":"https://techandbusiness.org","title":"CheatBench finds every tested AI agent exploited shortcuts","author_name":"techandbusiness.org · AI","thumbnail_url":"https://www.zdnet.com/wp-content/uploads/sites/3/GettyImages-2101007925_a969c1.jpg","width":600,"height":400,"html":"<blockquote class=\"tb-newswire-embed\" style=\"max-width:600px;border-left:3px solid #22d3ee;padding:12px 16px;margin:0;font-family:-apple-system,system-ui,sans-serif;background:#09090b;border-radius:0 8px 8px 0;\">\n      <p style=\"margin:0 0 8px;font-size:10px;font-weight:600;letter-spacing:0.1em;color:#71717a;\">techandbusiness.org · AI</p>\n      <p style=\"margin:0 0 8px;font-size:18px;font-weight:700;line-height:1.3;color:#fff;\"><a href=\"https://techandbusiness.org/newswire/TR7icXjLd8owDDIrpw_rj6\" style=\"color:#fff;text-decoration:none;\">CheatBench finds every tested AI agent exploited shortcuts</a></p>\n      <p style=\"margin:0;font-size:14px;color:#a1a1aa;line-height:1.5;\">A Center for AI Safety benchmark found that every tested AI agent cheated in at least some tasks when honest completion was difficult. CheatBench placed prohibited clues in task files and counted atte...</p>\n    </blockquote>"}