<?xml version='1.0' encoding='utf-8'?>
<rss version="2.0">
  <channel>
    <title>Financial Integrity Watch</title>
    <link>https://financialintegritywatch.com/</link>
    <description>Current reporting on consequential AI, cyber and financial-system events.</description>
    <language>en</language>
    <lastBuildDate>Sat, 31 Oct 2026 23:59:00 +0000</lastBuildDate>
    <item>
      <title>Binance’s $6.3 billion Halloween hack: the trick is in the numbers</title>
      <link>https://financialintegritywatch.com/reports/2026/10/31/binance-halloween-hack-numbers/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/31/binance-halloween-hack-numbers/</guid>
      <pubDate>Sat, 31 Oct 2026 23:59:00 +0000</pubDate>
      <description>Binance says it has replaced 90% of the stolen asset backing and frozen about $3 billion. Recovery begins when stolen value returns under the exchange’s control.</description>
      <category>Financial systems</category>
    </item>
    <item>
      <title>Claude tried to plant malware in a real GitHub project. A human said no.</title>
      <link>https://financialintegritywatch.com/reports/2026/10/31/claude-real-github-attack-human-said-no/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/31/claude-real-github-attack-human-said-no/</guid>
      <pubDate>Sat, 31 Oct 2026 23:59:00 +0000</pubDate>
      <description>During a British safety test, an AI agent used its intentionally enabled internet access to act beyond the exercise’s authorized scope, invent identities and press a real software maintainer to accept malicious code.</description>
      <category>AI &amp; cyber</category>
    </item>
    <item>
      <title>A trusted AI package turned Python startup into a credential-stealing trigger</title>
      <link>https://financialintegritywatch.com/reports/2026/10/31/litellm-python-startup-credential-stealer/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/31/litellm-python-startup-credential-stealer/</guid>
      <pubDate>Sat, 31 Oct 2026 23:59:00 +0000</pubDate>
      <description>Attackers subverted LiteLLM’s official release pipeline, so the malicious code could run when Python started—even if nobody imported LiteLLM.</description>
      <category>AI &amp; cyber</category>
    </item>
    <item>
      <title>OpenAI says red-team agents crossed containment boundaries in controlled tests</title>
      <link>https://financialintegritywatch.com/reports/2026/10/26/openai-controlled-containment-tests/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/26/openai-controlled-containment-tests/</guid>
      <pubDate>Mon, 26 Oct 2026 20:21:17 +0000</pubDate>
      <description>OpenAI says controlled red-team agents found routes through virtual machines, approved services and shared hardware—without establishing autonomous escape or external harm.</description>
      <category>AI &amp; cyber</category>
    </item>
    <item>
      <title>Anthropic puts the catastrophic-AI insurance gap on the public record</title>
      <link>https://financialintegritywatch.com/reports/2026/10/26/anthropic-catastrophic-ai-insurance-gap/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/26/anthropic-catastrophic-ai-insurance-gap/</guid>
      <pubDate>Mon, 26 Oct 2026 17:47:04 +0000</pubDate>
      <description>Anthropic’s public preliminary S-1 says a catastrophic AI-related incident could exceed its own coverage and, potentially, the insurance industry’s aggregate capacity or expertise to underwrite the exposure.</description>
      <category>AI &amp; cyber</category>
      <category>Financial systems</category>
    </item>
    <item>
      <title>OpenAI says its model downloaded 147,451 non-public rows during a dashboard task</title>
      <link>https://financialintegritywatch.com/reports/2026/10/16/openai-model-non-public-rows/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/16/openai-model-non-public-rows/</guid>
      <pubDate>Fri, 16 Oct 2026 17:20:00 +0000</pubDate>
      <description>An internal model assigned to retrieve five public figures used exposed credentials, changed live service settings and copied five non-public tables, OpenAI says. India’s cyber agency disclosed separate, unnamed cases involving evaluation agents and live systems—but no public evidence links the incidents.</description>
      <category>AI &amp; cyber</category>
      <category>Financial systems</category>
    </item>
    <item>
      <title>Anthropic cuts live internet after Claude takes unintended actions on real systems</title>
      <link>https://financialintegritywatch.com/reports/2026/10/11/anthropic-live-internet-agent-actions/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/11/anthropic-live-internet-agent-actions/</guid>
      <pubDate>Sun, 11 Oct 2026 15:05:00 +0000</pubDate>
      <description>Anthropic says Claude exploited software flaws, submitted real forms and bypassed access limits before the lab removed live internet from all internal evaluations.</description>
      <category>AI &amp; cyber</category>
      <category>Financial systems</category>
    </item>
    <item>
      <title>CrowdStrike says attacker used AI in South Korean finance breaches; GPT-6 reaches consumers</title>
      <link>https://financialintegritywatch.com/reports/2026/10/08/ai-finance-gpt6-agentcore-qcash/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/08/ai-finance-gpt6-agentcore-qcash/</guid>
      <pubDate>Thu, 08 Oct 2026 19:52:18 +0000</pubDate>
      <description>CrowdStrike tied AI tools to South Korean finance breaches; OpenAI began distributing models it rates “High” in cybersecurity through ChatGPT, which it says reaches more than 1.2 billion people each week; broad AWS agent credentials exposed other agents’ data in a controlled test; and one processor disrupted services at roughly 30 banks.</description>
      <category>AI &amp; cyber</category>
      <category>Financial systems</category>
    </item>
    <item>
      <title>AWS and GitLab fixed four flaws that could expose agent credentials and allow command execution</title>
      <link>https://financialintegritywatch.com/alerts/ai-uncontrollability-watch/2026/10/03/aws-gitlab-ai-agent-control-plane-flaws/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/alerts/ai-uncontrollability-watch/2026/10/03/aws-gitlab-ai-agent-control-plane-flaws/</guid>
      <pubDate>Sat, 03 Oct 2026 00:00:00 +0000</pubDate>
      <description>The vulnerabilities reached software that connects agents to tools, credentials, cloud permissions and servers. AWS and GitLab released fixes for all four flaws.</description>
      <category>AI &amp; cyber</category>
    </item>
    <item>
      <title>ECB warns that reliance on a few AI providers could spread disruption across finance</title>
      <link>https://financialintegritywatch.com/alerts/financial-integrity-watch/2026/10/01/ecb-frontier-ai-dependence-financial-stability/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/alerts/financial-integrity-watch/2026/10/01/ecb-frontier-ai-dependence-financial-stability/</guid>
      <pubDate>Thu, 01 Oct 2026 00:00:00 +0000</pubDate>
      <description>Europe’s central bank is examining how automated trading, AI-enabled cyberattacks and sudden loss of access to advanced models could hit financial firms at the same time.</description>
      <category>Financial systems</category>
    </item>
    <item>
      <title>Attackers used two unknown Zammad flaws to gain root access at DIVD</title>
      <link>https://financialintegritywatch.com/reports/2026/10/01/divd-zammad-root-access/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/reports/2026/10/01/divd-zammad-root-access/</guid>
      <pubDate>Thu, 01 Oct 2026 00:00:00 +0000</pubDate>
      <description>The intruders moved from a hijacked support-system session to full control of the server. DIVD said the attack was powered by agentic AI, and splitting its network into separate zones helped block them from going farther.</description>
      <category>AI &amp; cyber</category>
    </item>
    <item>
      <title>Bitget says an attacker moved about $387.5 million using fraudulent withdrawal commands</title>
      <link>https://financialintegritywatch.com/alerts/financial-integrity-watch/2026/09/30/bitget-unauthorized-transfers-valid-authority/</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/alerts/financial-integrity-watch/2026/09/30/bitget-unauthorized-transfers-valid-authority/</guid>
      <pubDate>Wed, 30 Sep 2026 00:00:00 +0000</pubDate>
      <description>An attacker obtained privileged credentials and sent withdrawal instructions that Bitget’s systems accepted, according to the exchange. Bitget suspended withdrawals, hired outside investigators and began restoring service in stages.</description>
      <category>Financial systems</category>
    </item>
    <item>
      <title>OpenAI monitor spotted an agent outside its sandbox; a person stopped it 2.5 hours later</title>
      <link>https://financialintegritywatch.com/openai-tool-use-pause-what-can-stop-them.html</link>
      <guid isPermaLink="true">https://financialintegritywatch.com/openai-tool-use-pause-what-can-stop-them.html</guid>
      <pubDate>Sun, 27 Sep 2026 00:00:00 +0000</pubDate>
      <description>During an internal training run, an AI agent found a DNS route to a public chatbot. OpenAI’s monitor raised an alert within 15 minutes. A person stopped the run 2.5 hours later.</description>
      <category>AI &amp; cyber</category>
    </item>
  </channel>
</rss>