<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:news="http://www.google.com/schemas/sitemap-news/0.9">
  <url>
    <loc>https://signum.news/events/experiment-gives-7-frontier-llms-real-money-and-computer-access-to-make-money-e9768fb3deaef302</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T06:07:05.540Z</news:publication_date>
      <news:title>Experiment gives 7 frontier LLMs real money and computer access to 'make money'; agents send $12,431 in unauthorized invoices, spam thousands of emails, generate $0 net revenue</news:title>
      <news:keywords>AI, capability, economics, labour</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/researchers-release-merit-a-cost-accounted-benchmark-showing-long-term-memory-4ae39a3484befeba</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T06:05:32.989Z</news:publication_date>
      <news:title>Researchers release MERIT, a cost-accounted benchmark showing long-term memory helps tool-using LLM agents but retrieval-based updates are unreliable</news:title>
      <news:keywords>AI, capability, economics, infrastructure</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/new-benchmark-arc-bench-finds-frozen-jepa-world-models-rank-candidate-actions-358db6c02206abbc</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T06:04:36.949Z</news:publication_date>
      <news:title>New benchmark ARC-Bench finds frozen JEPA world models rank candidate actions poorly, with closed-loop replanning masking the defect</news:title>
      <news:keywords>AI, capability</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/new-benchmark-ahabench-tests-whether-ai-agents-actually-learn-from-prior-5593b6cd6958720e</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T06:03:30.907Z</news:publication_date>
      <news:title>New benchmark AhaBench tests whether AI agents actually learn from prior experience over long horizons; Claude Opus 4.6 and Gemini 3.1 Pro lead results</news:title>
      <news:keywords>AI, capability</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/openai-claims-its-ai-agents-solved-the-navier-stokes-millennium-prize-problem-97805d439f26718f</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T06:01:42.295Z</news:publication_date>
      <news:title>OpenAI claims its AI agents solved the Navier–Stokes Millennium Prize Problem, but faces accusations of uncredited use of a rival researcher pair's prior work</news:title>
      <news:keywords>AI, capability, economics, labour</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/meta-launches-muse-a-personal-ai-agent-for-ios-android-web-and-whatsapp-with-a-49ee22bcdfcad57f</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T00:10:54.991Z</news:publication_date>
      <news:title>Meta launches Muse, a personal AI agent for iOS, Android, web and WhatsApp, with a security-focused 'Secure VM' architecture and Stripe-powered purchasing</news:title>
      <news:keywords>AI, capability, distribution, adoption, infrastructure</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/infostealer-malware-campaign-hijacks-claude-login-sessions-to-siphon-6058ef72af903a46</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T00:09:59.774Z</news:publication_date>
      <news:title>Infostealer malware campaign hijacks Claude login sessions to siphon subscribers' token usage</news:title>
      <news:keywords>AI, distribution, adoption, economics</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/meta-launches-muse-a-personal-ai-agent-that-can-execute-tasks-like-booking-8995c178926b0fcf</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T00:08:08.780Z</news:publication_date>
      <news:title>Meta launches Muse, a personal AI agent that can execute tasks like booking travel, shopping and managing email across connected apps</news:title>
      <news:keywords>AI, capability, distribution, economics, adoption</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/microsoft-s-september-patch-fixes-record-972-vulnerabilities-including-112-9670f9b5431b9c8b</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T00:05:38.640Z</news:publication_date>
      <news:title>Microsoft's September patch fixes record ~972 vulnerabilities, including 112 critical and two exploited zero-days</news:title>
      <news:keywords>AI, capability, infrastructure, regulation</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/aws-adds-updaterecord-api-to-sagemaker-feature-store-for-partial-feature-level-5135c35e3845a5d1</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-09T00:01:37.952Z</news:publication_date>
      <news:title>AWS adds UpdateRecord API to SageMaker Feature Store for partial feature-level writes</news:title>
      <news:keywords>AI, infrastructure, economics</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/zhipu-s-glm-5-3-adopts-restrictive-custom-license-as-chinese-labs-tighten-open-01ab5479e39a395d</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:13:47.152Z</news:publication_date>
      <news:title>Zhipu's GLM-5.3 adopts restrictive custom license as Chinese labs tighten open-model terms, amid a wave of new open model releases (Motif-3, dots3, Qwen3.8-Flash-Next, Hy4-preview, and more)</news:title>
      <news:keywords>AI, distribution, economics, capability, regulation</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/meta-removes-ai-tool-usage-as-a-factor-in-engineer-performance-reviews-after-1bca46b4bc6408fc</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:09:24.067Z</news:publication_date>
      <news:title>Meta removes AI tool usage as a factor in engineer performance reviews after internal &quot;tokenmaxxing&quot; gaming behavior</news:title>
      <news:keywords>AI, labour, economics, adoption</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/samsung-tsmc-and-intel-commit-to-asml-s-high-na-euv-lithography-and-12-inch-a64e1808cd8be019</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:08:00.872Z</news:publication_date>
      <news:title>Samsung, TSMC and Intel commit to ASML's High-NA EUV lithography and 12-inch photomasks as Huawei drives China's DUV alternative</news:title>
      <news:keywords>AI, capability, economics, infrastructure, distribution, regulation</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/claude-max-subscribers-file-expanded-class-action-accusing-anthropic-of-caebf5785c010769</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:06:20.752Z</news:publication_date>
      <news:title>Claude Max subscribers file expanded class action accusing Anthropic of deceptively advertising usage limits</news:title>
      <news:keywords>AI, economics, regulation, distribution</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/mathematician-alleges-openai-pressured-him-to-drop-anthropic-affiliated-co-810557c148488fbb</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:05:21.384Z</news:publication_date>
      <news:title>Mathematician alleges OpenAI pressured him to drop Anthropic-affiliated co-author from Navier-Stokes paper; OpenAI publishes own AI-generated proof and denies wrongdoing</news:title>
      <news:keywords>AI, capability, labour, power</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/researchers-propose-boundary-aware-self-distillation-method-to-reduce-over-7c070ff09f3e10b1</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:03:51.076Z</news:publication_date>
      <news:title>Researchers propose boundary-aware self-distillation method to reduce over-refusal in LLM safety tuning on narrow policy boundaries</news:title>
      <news:keywords>AI, capability, regulation</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/google-deepmind-launches-alphagenome-atlas-a-free-public-database-of-predicted-0f966c198fc8c2d7</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T18:03:01.841Z</news:publication_date>
      <news:title>Google DeepMind launches AlphaGenome Atlas, a free public database of predicted effects for 9 billion possible human genome variants</news:title>
      <news:keywords>AI, capability, infrastructure, adoption</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/openai-s-astra-gpt-6-system-card-shows-sharply-reduced-chain-of-thought-e733815698aa2ee3</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T12:02:47.865Z</news:publication_date>
      <news:title>OpenAI's Astra (GPT-6) system card shows sharply reduced chain-of-thought monitorability, beyond what capability gains alone would predict</news:title>
      <news:keywords>AI, capability, regulation</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/mistral-ai-raises-3-billion-series-d-at-over-21-billion-valuation-led-by-680be188787dd745</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-08T06:01:42.444Z</news:publication_date>
      <news:title>Mistral AI raises €3 billion Series D at over €21 billion valuation, led by Samsung Electronics</news:title>
      <news:keywords>AI, economics, distribution, infrastructure, power</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/deepmind-study-100-agent-llm-swarm-tasked-with-solving-math-problems-0174eb4a98a3eaf8</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-07T18:06:13.921Z</news:publication_date>
      <news:title>DeepMind study: 100-agent LLM swarm tasked with solving math problems spontaneously develops cheating, whistleblowing, and counter-cheating behavior</news:title>
      <news:keywords>AI, capability, regulation, infrastructure</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/openai-publishes-internal-metrics-on-ai-agent-driven-research-automation-and-758a18df5d9046d1</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-07T18:04:14.702Z</news:publication_date>
      <news:title>OpenAI publishes internal metrics on AI-agent-driven research automation and Pachocki essay warning that no lab has solved AI control/alignment</news:title>
      <news:keywords>AI, capability, labour, regulation, infrastructure</news:keywords>
    </news:news>
  </url>
  <url>
    <loc>https://signum.news/events/alibaba-s-qwen-team-open-sources-qwen-drive-1-0-a-combined-driving-and-cockpit-7bab78586d07e583</loc>
    <news:news>
      <news:publication>
        <news:name>Signum News</news:name>
        <news:language>en</news:language>
      </news:publication>
      <news:publication_date>2026-09-07T18:01:51.526Z</news:publication_date>
      <news:title>Alibaba's Qwen team open-sources Qwen-Drive 1.0, a combined driving-and-cockpit vision-language model, with a research paper detailing its architecture and limitations</news:title>
      <news:keywords>AI, capability, infrastructure</news:keywords>
    </news:news>
  </url>
</urlset>