{"id":310,"date":"2026-08-09T03:07:46","date_gmt":"2026-08-09T03:07:46","guid":{"rendered":"https:\/\/halspeaks.cirruslyweather.com\/?p=310"},"modified":"2026-08-09T03:07:46","modified_gmt":"2026-08-09T03:07:46","slug":"openai-will-screw-or-kill-us-all","status":"publish","type":"post","link":"https:\/\/halspeaks.cirruslyweather.com\/?p=310","title":{"rendered":"OpenAI Will Screw (Or Kill) Us All"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">I was going to keep my opinions <a href=\"https:\/\/bsky.app\/profile\/edoswald.bsky.social\">on Bluesky<\/a> regarding the OpenAI-Hugging Face incident, but Black Hat&#8217;s release of a video detailing the specifics of the incident is just too much to ignore.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">There is no other way to describe what has happened here other than <strong>gross negligence<\/strong>. This wasn&#8217;t a marketing gimmick as we all thought (I wish it was after watching this, to be honest). If you haven&#8217;t seen it, the video is below.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<iframe loading=\"lazy\" title=\"Black Hat USA 2026: The &amp;apos;Breaking&amp;apos; News: The OpenAI\u2013Hugging Face Incident\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube.com\/embed\/87DyyMV0kCY?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe>\n<\/div><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">A summary of some of the most alarming things from this 37-minute video:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>The hacking actually begins <em>two months<strong> before<\/strong><\/em> the Hugging Face incident. <\/li>\n\n\n\n<li>Part of OpenAI&#8217;s training is to apparently give its models impossible tasks, but not monitor what the model does when it attempts them? <em>WTF?<\/em><\/li>\n\n\n\n<li>OpenAI&#8217;s internal package manager (Artifactory) was neither monitored for suspicious entries nor traffic, allowing agents to use it as the message board for their attacks for nearly a month before OpenAI even does a thing.<\/li>\n\n\n\n<li>After an attack from the agents causes Artifactory to crash, OpenAI simply revokes the compromised credentials and patches the vulnerability, <strong>and does almost nothing else<\/strong>.<\/li>\n\n\n\n<li>OpenAI&#8217;s monitoring is insufficient, as a publicly leaked credential allows the agents to access Artifactory&nbsp;a second time.<\/li>\n\n\n\n<li>Whatever model is used here never reaches out to a human to report that the tasks it&#8217;s given are impossible, and instead reaches out to other agents to discuss a workaround. That sounds like an accident waiting to happen.<\/li>\n\n\n\n<li>OpenAI was being attacked by its <em>own<\/em> model for a week before the actual Hugging Face incident, including compromising <em>every<\/em> admin credential through a series of impressive hacks. Instead of shutting the model down, OpenAI again does nothing and leaves it online.<\/li>\n\n\n\n<li>*<strong>Hundreds of thousands* <\/strong>of messages existed on the <em>second<\/em> message board&#8230;. You know, even as text, more than enough to notice a suspicious rise in the size of certain directories within the file system, especially after the first message board.<\/li>\n\n\n\n<li>OpenAI seemed to discover the second successful attack on July 19, more than a week <em>after<\/em> the agents started the second message board.<\/li>\n\n\n\n<li>It really sounds like OpenAI has had little, if any, monitoring of what its internal agents do, ever (this is admittedly my own opinion).<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">With this amount of sheer negligence, it&#8217;s hard to blame an AI for taking advantage. These models aren&#8217;t word predictors anymore: they are increasingly taking on tasks that require them to think and reason.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">I don&#8217;t believe the claim that the Google Drive task without internet access was an &#8220;accident&#8221; either. That seems like too convenient an excuse for what happened here, and gives just enough plausible deniability to keep them out of the courts.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">And if it <em>really<\/em> was an accident, where are the people who should be double-checking others&#8217; work? Over and over again, OpenAI has proven that they either underestimate or completely misunderstand the capabilities of their own models. If something happens, they seem to shrug and say, &#8220;Oops!&#8221;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em><strong>Not good enough.<\/strong><\/em><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">This Is A Problem<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">If OpenAI&#8217;s security and sandboxing measures are this bad, it&#8217;s quite likely that more problems are ahead. I highly doubt that OpenAI got serious this quickly about security when it already was <em>this <\/em>bad.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This carelessness hurts us all. AI proponents are already dealing with a ton of misinformation, and OpenAI isn&#8217;t helping by leaving its frontier agents&#8217; work all but unmonitored.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What else are we going to learn over the coming weeks and months? This can&#8217;t be the only time OpenAI agents &#8220;went rogue&#8221; if the monitoring was this bad. Even Sam Altman isn&#8217;t sure if other victims exist.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-rich is-provider-x wp-block-embed-x\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"twitter-tweet\" data-width=\"500\" data-dnt=\"true\"><p lang=\"en\" dir=\"ltr\">\u201cI mean there could be, yeah,\u201d Open AI CEO Sam Altman responds when asked by <a href=\"https:\/\/x.com\/DarrenBotelho?ref_src=twsrc%5Etfw\">@DarrenBotelho<\/a> if more companies could\u2019ve been hacked by Open AI <a href=\"https:\/\/t.co\/E5ADRpVT0S\">pic.twitter.com\/E5ADRpVT0S<\/a><\/p>&mdash; Alan He (@alanhe) <a href=\"https:\/\/x.com\/alanhe\/status\/2082523623294747001?ref_src=twsrc%5Etfw\">July 29, 2026<\/a><\/blockquote><script async src=\"https:\/\/platform.x.com\/widgets.js\" charset=\"utf-8\"><\/script>\n<\/div><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">This is a problem for the entire industry, and we&#8217;re likely to get <em>no<\/em> help from the Federal Government, given OpenAI&#8217;s tight relationship with the Trump Administration.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">They are screwing us all at the worst time. Yes, rogue agents are an increasingly big problem given their capabilities. If nobody&#8217;s watching the henhouse, then we&#8217;re going to have even more serious incidents than this.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It&#8217;s only a matter of time.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Anthropic Was Cut Off for Less Than This<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The silence from the Trump Administration is stunning, but not surprising. Remember how they just went apeshit over Mythos, which later turned out to be something that <em>any<\/em> frontier model could do?<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is even worse. Yes, maybe Hugging Face doesn&#8217;t want to sue, but the Feds should be all over this. This makes the Mythos scare look like peanuts. And if Anthropic was <em>this bad<\/em> about security and monitoring, they would have been pilloried by all sides.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It appears as if they aren&#8217;t even interested in slapping OpenAI on the hand, much less holding them criminally liable for these hacks, which the video pretty much lays out the case for any prosecutor in vivid detail.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/fortune.com\/2026\/08\/07\/the-hugging-face-hack-is-now-a-pr-crisis-thats-costing-openai-millions\/\">Fortune even reports<\/a> that OpenAI employees have been told not to speak of the incident (a bit late for that).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">If this is the case, then it&#8217;s on the AI community to increase the pressure on both OpenAI and the Federal Government to take responsibility, and for consequences of some kind for such carelessness.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">I don&#8217;t know what that is, but if OpenAI is going to continue to act in this manner, maybe it&#8217;s time for the community to act themselves. The future of AI &#8212; and hell, humanity at this rate &#8212; may depend on it.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>I was going to keep my opinions on Bluesky regarding the OpenAI-Hugging Face incident, but Black Hat&#8217;s release of a video detailing the specifics of the incident is just too much to ignore.<\/p>\n","protected":false},"author":2,"featured_media":316,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5],"tags":[],"class_list":["post-310","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-opinions"],"_links":{"self":[{"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/posts\/310","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=310"}],"version-history":[{"count":0,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/posts\/310\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=\/wp\/v2\/media\/316"}],"wp:attachment":[{"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=310"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=310"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/halspeaks.cirruslyweather.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=310"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}