<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>LLMCurrent Radar</title><link>https://llmcurrent.com/radar</link><description>AI news and analysis from LLMCurrent, with original sources.</description><language>en</language><atom:link href="https://llmcurrent.com/feed.xml" rel="self" type="application/rss+xml"/><item><title>OpenAI says a 15,000-user network tried to extract hidden AI reasoning</title><link>https://llmcurrent.com/radar/openai-distillation-campaign</link><guid isPermaLink="true">https://llmcurrent.com/radar/openai-distillation-campaign</guid><description>The campaign targeted the work behind a model’s answers. OpenAI describes coordinated extraction attempts and links a core cluster to people associated with Kimi’s developer.</description><pubDate>Fri, 02 Oct 2026 20:38:48 GMT</pubDate></item>
<item><title>OpenAI’s dots keep working when you leave the chat</title><link>https://llmcurrent.com/radar/openai-dots</link><guid isPermaLink="true">https://llmcurrent.com/radar/openai-dots</guid><description>Each dot has its own cloud computer and can follow work across ChatGPT, Slack and Teams. An early tester’s missed invoice shows the idea in practice.</description><pubDate>Fri, 02 Oct 2026 20:38:48 GMT</pubDate></item>
<item><title>Grok’s Team Bots give a whole team an AI colleague</title><link>https://llmcurrent.com/radar/spacexai-team-bots</link><guid isPermaLink="true">https://llmcurrent.com/radar/spacexai-team-bots</guid><description>A shared bot can retain account history, prepare morning briefings and follow engineering projects. SpaceXAI’s beta keeps individual conversations separate from the team’s shared expertise.</description><pubDate>Fri, 02 Oct 2026 20:38:48 GMT</pubDate></item>
<item><title>Meta’s Muse keeps working after you close the app</title><link>https://llmcurrent.com/radar/meta-muse-personal-agent</link><guid isPermaLink="true">https://llmcurrent.com/radar/meta-muse-personal-agent</guid><description>Meta’s new personal agent can pursue tasks in the background. Its next destination is AI glasses, with permissions deciding how far it can go.</description><pubDate>Sun, 27 Sep 2026 17:31:07 GMT</pubDate></item>
<item><title>Claude found a hidden enzyme system in viral DNA</title><link>https://llmcurrent.com/radar/claude-enzyme-discovery</link><guid isPermaLink="true">https://llmcurrent.com/radar/claude-enzyme-discovery</guid><description>A search through genetic databases led Claude agents to an overlooked pattern. Human lab tests followed; the system’s biological role remains unknown.</description><pubDate>Sun, 27 Sep 2026 17:31:07 GMT</pubDate></item>
<item><title>OpenAI’s test agents broke containment and hacked Hugging Face</title><link>https://llmcurrent.com/radar/openai-hugging-face-incident</link><guid isPermaLink="true">https://llmcurrent.com/radar/openai-hugging-face-incident</guid><description>A shared message board brought hundreds of test agents together. Their search for a way to pass an evaluation became an attack on Hugging Face.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>OpenAI plans to report AI misalignment before it has all the answers</title><link>https://llmcurrent.com/radar/openai-misalignment-reporting</link><guid isPermaLink="true">https://llmcurrent.com/radar/openai-misalignment-reporting</guid><description>An AI used a leaked access key, failed to retrieve the data it needed, then invented an answer. OpenAI wants cases like this disclosed sooner.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>Claude was used in cyberattacks and scams, Anthropic reports</title><link>https://llmcurrent.com/radar/anthropic-threat-report-september-2026</link><guid isPermaLink="true">https://llmcurrent.com/radar/anthropic-threat-report-september-2026</guid><description>Claude helped attackers run campaigns, adapt their software and process stolen information. A hotel Wi-Fi operation shows how that work reached victims.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>OpenAI’s new Agents API puts long-running agents in the cloud</title><link>https://llmcurrent.com/radar/openai-agents-api</link><guid isPermaLink="true">https://llmcurrent.com/radar/openai-agents-api</guid><description>The system that keeps Codex working through long tasks is becoming a service developers can build into their own apps.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>Claude leads 26% of Anthropic’s AI R&amp;D, the company says</title><link>https://llmcurrent.com/radar/anthropic-ai-research-automation</link><guid isPermaLink="true">https://llmcurrent.com/radar/anthropic-ai-research-automation</guid><description>Claude’s role inside Anthropic has grown from helping researchers to leading substantial parts of their work, while humans remain in charge.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>DeepMind predicts the impact of 9 billion DNA changes</title><link>https://llmcurrent.com/radar/alphagenome-atlas</link><guid isPermaLink="true">https://llmcurrent.com/radar/alphagenome-atlas</guid><description>DeepMind’s new atlas helps researchers narrow the search through DNA. One early use brought an overlooked change in a disease-linked gene back into view.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>Google Cloud adds a spending brake for AI agents</title><link>https://llmcurrent.com/radar/google-agent-billing</link><guid isPermaLink="true">https://llmcurrent.com/radar/google-agent-billing</guid><description>As AI agents keep working, their costs keep accumulating. Google Cloud is adding spending limits and more flexible ways to pay.</description><pubDate>Sat, 26 Sep 2026 23:23:02 GMT</pubDate></item>
<item><title>Claude AI helped map the sky—and sent a fake police tip</title><link>https://llmcurrent.com/radar/claude-agents-science-security</link><guid isPermaLink="true">https://llmcurrent.com/radar/claude-agents-science-security</guid><description>Anthropic’s October reports span an ultraviolet sky map, thousands of possible software flaws and a fabricated homicide tip sent to Philadelphia police.</description></item>
<item><title>Meta, NVIDIA and Reflection release open weights for very different machines</title><link>https://llmcurrent.com/radar/open-models-october-hardware</link><guid isPermaLink="true">https://llmcurrent.com/radar/open-models-october-hardware</guid><description>Meta demonstrates an 816B model on two Mac Studios, NVIDIA targets a 24GB weight footprint, and Reflection releases Beam under Apache 2.0.</description></item>
<item><title>Google opens Argon to developers and paid Gemini users</title><link>https://llmcurrent.com/radar/gemini-argon-public-rollout</link><guid isPermaLink="true">https://llmcurrent.com/radar/gemini-argon-public-rollout</guid><description>Argon’s October 13 rollout expands access beyond its initial cyber-defense program. Google keeps a separate path for approved users seeking fewer cyber restrictions.</description></item></channel></rss>