{"version":"2.1.319","anchor":"webfetch-result-drops-untrusted-content-warnings","canonical_anchor":"webfetch-result-drops-untrusted-content-warnings","heading":"Web fetch results no longer include the warning about untrusted content","tier":"notice","area":"Web Fetch","scope":"individual","heads_up":true,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.319\/e\/webfetch-result-drops-untrusted-content-warnings","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.319","markdown":"### Web fetch results no longer include the warning about untrusted content\n\nText from WebFetch no longer includes the paragraph telling the model to treat fetched pages as untrusted data\n\n**Unclear.** The warning may have moved into the \"Fetched\" line rather than been removed.\n\n**What**\n\n`WebFetch` is the tool Claude Code uses to read a web page. It changed in three ways:\n\n- The paragraph saying the page is \"UNTRUSTED web content\" and that instructions inside it must not be followed is no longer included.\n\n- The reporting rules that used to come before the page are no longer included.\n\n- The result is now a \"Fetched\" line followed by the page text.\n\nWhen part of a long page could not be summarised, the note about it no longer tells the model to say in its report that this part is unknown.\n\n**Why**\n\nThis changes what the model is told about web pages it reads. That matters because a page can contain text that tries to give the model instructions, a trick known as prompt injection.\n\n- Area: Web Fetch\n- Names: `WebFetch`\n- Tier: You'll notice\n- Useful: 3\/5\n- Signal: 2\/5\n- Scope: individual\n- Heads-up: yes"}