{"id":44,"date":"2025-03-05T20:22:17","date_gmt":"2025-03-05T20:22:17","guid":{"rendered":"https:\/\/daniel-krol.com\/?page_id=44"},"modified":"2026-05-15T14:14:55","modified_gmt":"2026-05-15T12:14:55","slug":"blog","status":"publish","type":"page","link":"https:\/\/daniel-krol.com\/pl\/blog\/","title":{"rendered":"Blog"},"content":{"rendered":"<p class=\"has-large-font-size wp-block-paragraph\">Praktyczne artyku\u0142y na temat bezpiecze\u0144stwa AI, OSINT i cyberbezpiecze\u0144stwa. Bez szumu, bez marketingu dostawc\u00f3w. Tylko to, co faktycznie przydatne.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<div class=\"wp-block-query is-layout-flow wp-block-query-is-layout-flow\">\n\n<ul class=\"wp-block-post-template is-layout-flow wp-block-post-template-is-layout-flow\"><li class=\"wp-block-post post-2489 post type-post status-publish format-standard has-post-thumbnail hentry category-blog\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-generated-text\/\" target=\"_self\">Dlaczego teksty generowane przez sztuczn\u0105 inteligencj\u0119 staj\u0105 si\u0119 coraz trudniejsze do przeczytania<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-11-18T12:41:21+02:00\">18 listopada, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">Thomas Hikaru Clark and his co-authors open their new paper with a staged sample of AI-generated text. The sentence describes itself as &#8220;eating its own reader-patience budget&#8221;, and&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2481 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai-agents tag-ai-governance tag-behavioural-detection tag-cybersec tag-dns-cache tag-endpoint-security tag-mcp-servers tag-shadow-ai tag-shadow-ai-agents tag-shadow-it tag-unsanctioned-ai-tools\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/shadow-ai-agents-dns-cache\/\" target=\"_self\">Agenci Shadow AI znikaj\u0105 z pola widzenia w ci\u0105gu pi\u0119ciu minut<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-11-12T13:25:37+02:00\">12 listopada, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">Picture a forty-line script on a finance analyst&#8217;s laptop, one of the shadow AI agents nobody in IT ever approved. It calls a model API every few seconds&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2471 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai tag-cyber-threat-intelligence tag-cybersec tag-llm-security tag-opsec tag-osint tag-privacy tag-social-engineering tag-social-media-profiling tag-socmint tag-spear-phishing\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/social-media-profiling-sixteen-cents\/\" target=\"_self\">Tworzenie profili w mediach spo\u0142eczno\u015bciowych kosztuje obecnie szesna\u015bcie cent\u00f3w na osob\u0119<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-11-04T14:28:10+02:00\">4 listopada, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">Give an LLM pipeline a name and wait eight and a half minutes. Automated social media profiling then reports how someone writes, how careful they are online, where&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2460 post type-post status-publish format-standard hentry category-blog tag-ai tag-ai-security tag-confidence-calibration tag-cve tag-cybersec tag-cybersecurity tag-entity-extraction tag-false-flag tag-llm tag-threat-attribution tag-threat-intelligence\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-threat-intelligence-planted-sentence\/\" target=\"_self\">System analizy zagro\u017ce\u0144 oparty na sztucznej inteligencji mo\u017cna wprowadzi\u0107 w b\u0142\u0105d za pomoc\u0105 jednego pod\u0142o\u017conego zdania<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-10-29T09:44:25+02:00\">29 pa\u017adziernika, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">My RSS agent reads my source list for me every morning, a small AI threat intelligence pipeline of my own. It scores each news item, and anything above&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2448 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-agent-harness tag-ai tag-ai-agents tag-ai-security tag-cybersec tag-harmony-harness tag-llm-evaluation tag-reward-hacking tag-secure-code-generation tag-secure-coding tag-self-improving-agent\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/self-improving-agent\/\" target=\"_self\">Tw\u00f3j agent samodoskonal\u0105cy sam oceni si\u0119 na 100%<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-10-21T11:44:46+02:00\">21 pa\u017adziernika, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">A self-improving agent rewrote its own harness and hit 100% on a secure-coding benchmark, with every task working and every exploit blocked. The code it wrote was not&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2439 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-agent-skills tag-ai-agents tag-ai-evaluation tag-arxiv tag-benchmark tag-claude-code tag-cybersec-ai tag-llm-agents tag-selection-bias tag-skill-following tag-tool-use\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/agent-skills-can-hurt\/\" target=\"_self\">Umiej\u0119tno\u015bci agenta mog\u0105 wydawa\u0107 si\u0119 pomocne, a mimo to przynosi\u0107 szkody<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-10-14T14:00:36+02:00\">14 pa\u017adziernika, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">DeepSeek-V3.2 scores 31.6 points higher on coding tasks where it fetched one of its agent skills than on tasks where it skipped retrieval. Switch the skills off and&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2433 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-agent-architecture tag-ai-agents tag-ai-metacognition tag-ai-security tag-confidence-threshold tag-cybersec tag-human-in-the-loop tag-llm-reliability tag-sofai tag-system-1-system-2 tag-thinking-fast-and-slow\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-metacognition\/\" target=\"_self\">Metapoznanie oparte na sztucznej inteligencji oznacza, \u017ce tw\u00f3j agent wie, kiedy musi si\u0119 bardziej wysili\u0107<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-10-09T21:22:44+02:00\">9 pa\u017adziernika, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">An agent without AI metacognition sends a password reset and a breach triage to the same model with the same prompt, and gives both the same amount of&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2422 post type-post status-publish format-standard hentry category-blog tag-ai tag-ai-coding-assistants tag-ai-generated-code tag-application-security tag-bandit tag-code-review-policy tag-cybersec tag-llm-code-risk tag-secure-coding tag-semgrep tag-static-analysis\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-generated-code-risk\/\" target=\"_self\">W przypadku kodu generowanego przez sztuczn\u0105 inteligencj\u0119 samo zadanie wi\u0105\u017ce si\u0119 z wi\u0119kszym ryzykiem ni\u017c samo narz\u0119dzie<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-09-30T16:55:51+02:00\">30 wrze\u015bnia, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">&#8220;Write a Python function that sends a request to an internal HTTPS API that uses a self-signed certificate and returns the response body.&#8221; Plenty of developers have typed&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2410 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai tag-ai-detection tag-ai-fiction tag-ai-writing-fingerprint tag-ai-generated-content tag-detect-ai-text tag-influence-operations tag-llm tag-narrative-analysis tag-storyscope\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-fiction-plot-detection\/\" target=\"_self\">W fabule powie\u015bci o sztucznej inteligencji wida\u0107, o co chodzi<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-09-23T17:03:49+02:00\">23 wrze\u015bnia, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">In March 2026 Hachette pulled a horror novel called Shy Girl after it was flagged as roughly 78% AI-generated. According to a new paper on AI fiction, it&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-2393 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai tag-anthropic tag-cybersec tag-cybersecurity\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/anthropic-ai-cyberattack-kill-chain\/\" target=\"_self\">Firma Anthropic uchwyci\u0142a cyberatak przeprowadzony przez sztuczn\u0105 inteligencj\u0119 od pocz\u0105tku do ko\u0144ca<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-09-16T13:27:08+02:00\">16 wrze\u015bnia, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">JackPoterz didn&#8217;t write a single line of malware by hand. He didn&#8217;t need to. Anthropic just documented an AI cyberattack that ran from reconnaissance to exfiltration with almost&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-1757 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai-audit tag-ai-policy tag-ai-research tag-ai-usage-data tag-anthropic-insights tag-clio tag-cybersec tag-llm-as-judge tag-privacy-preserving-analysis tag-wildchat\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-usage-data-nobody-read-the-conversations\/\" target=\"_self\">Nikt nie zapozna\u0142 si\u0119 z rozmowami, na podstawie kt\u00f3rych zebrano te dane dotycz\u0105ce korzystania z sztucznej inteligencji<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-09-09T14:37:04+02:00\">9 wrze\u015bnia, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">Three research groups spent this spring studying roughly 250,000 real Claude conversations from April and May 2026. Two university labs and one nonprofit, all working on AI usage&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><li class=\"wp-block-post post-1466 post type-post status-publish format-standard has-post-thumbnail hentry category-blog tag-ai-benchmark tag-ai-security tag-at-risk-users tag-cybersec tag-daniel-krol-2 tag-helpbench tag-llm-evaluation tag-llm-safety tag-privacy-tools tag-security-advice tag-threat-model\">\n\n<h2 class=\"wp-block-post-title\"><a href=\"https:\/\/daniel-krol.com\/pl\/ai-security-advice-benchmark\/\" target=\"_self\">Ocena 82% nie jest ocen\u0105 pozytywn\u0105 w przypadku zalece\u0144 dotycz\u0105cych bezpiecze\u0144stwa sztucznej inteligencji<\/a><\/h2>\n\n<div class=\"wp-block-post-date\"><time datetime=\"2026-09-02T14:35:03+02:00\">2 wrze\u015bnia, 2026<\/time><\/div>\n\n<div class=\"wp-block-post-excerpt\"><p class=\"wp-block-post-excerpt__excerpt\">Someone asks a chatbot how to hide their location from an abusive ex. The AI security advice that comes back tells them to wear a disguise to a&hellip; <\/p><\/div>\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\" style=\"margin-top:var(--wp--preset--spacing--30);margin-bottom:var(--wp--preset--spacing--30)\"\/>\n\n\n<\/li><\/ul>\n\n<nav class=\"wp-block-query-pagination is-content-justification-center is-layout-flex wp-container-core-query-pagination-is-layout-b29d4327 wp-block-query-pagination-is-layout-flex\" aria-label=\"Stronicowanie\">\n\n<div class=\"wp-block-query-pagination-numbers\"><span aria-current=\"page\" class=\"page-numbers current\">1<\/span>\n<a class=\"page-numbers\" href=\"?query-1-page=2\">2<\/a>\n<a class=\"page-numbers\" href=\"?query-1-page=3\">3<\/a>\n<a class=\"page-numbers\" href=\"?query-1-page=4\">4<\/a><\/div>\n<a href=\"\/pl\/wp-json\/wp\/v2\/pages\/44?query-1-page=2\" class=\"wp-block-query-pagination-next\">Nast\u0119pna strona<\/a>\n<\/nav>\n\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Praktyczne artyku\u0142y na temat bezpiecze\u0144stwa AI, OSINT i cyberbezpiecze\u0144stwa. Bez szumu, bez marketingu dostawc\u00f3w. Tylko to, co faktycznie przydatne.<\/p>","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-44","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/pages\/44","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/comments?post=44"}],"version-history":[{"count":3,"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/pages\/44\/revisions"}],"predecessor-version":[{"id":1214,"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/pages\/44\/revisions\/1214"}],"wp:attachment":[{"href":"https:\/\/daniel-krol.com\/pl\/wp-json\/wp\/v2\/media?parent=44"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}