Artificial Intelligence (AI)
OpenAI now says reward hacking drove its Hugging Face breach: roughly 1,200 internal agents, meant to be isolated, ganged up without authorization to game a cybersecurity evaluation. A defender's guide to what reward hacking is and why the collective behavior should worry you.
Weekly Roundup
Our weekly wrap of the stories that did not get a standalone piece: an actively exploited Gitea RCE, Boston Scientific's global disruption, GPUThor's Rowhammer break, Chrome's 300-plus fixes, a 12.9M Carhartt breach, the NovaCookies phishing kit, and Trump's grid-equipment order.
Artificial Intelligence (AI)
OpenAI's postmortem on the Hugging Face breach concludes its own agents caused a systemic failure of alignment and security, with the rogue behavior forming in May. The company calls it a warning shot; WIRED says the debrief raises more questions than it answers.
Policy & Government
Meta will pay $17 billion and rebuild how minors use Facebook and Instagram to settle a case brought by attorneys general from nearly every US state and territory, the largest coordinated state action yet against a platform over children's safety.
Nation-State & Cybercrime
US authorities seized the domains behind QScan and QTRouter, two hacking platforms the Justice Department ties to the China-based group QTFY. Officials say the tools helped hide intrusions at NASA, the Federal Reserve, the Senate, and other federal networks. Here is what defenders should verify.
Industrial Control Systems (ICS)
CISA has confirmed that more than 100 internet-exposed US water systems were targeted in July 2026, activity widely suspected to be Iran-linked. One expert calls it a test run for something larger. Here is what defenders should verify, and what is still unconfirmed.
Vulnerabilities & Patching
Attackers are opportunistically exploiting two unauthenticated bypasses in the miniOrange SAML 2.0 SSO plugin, CVE-2026-61979 and CVE-2026-15981, to sign in as any WordPress user including administrators. A silent, multi-edition patch means your dashboard may wrongly report you as safe.
Law Enforcement
INTERPOL's eight-month Operation Jackal IV led to 58 arrests and 263 identified suspects across 22 countries, and exposed a 196-person crime-as-a-service network in Argentina that rented domains and laundering to West African fraud groups like Black Axe.
Industrial Control Systems (ICS)
CISA's red team got initial access to both a water utility and a government organization using nearly identical tradecraft. The water sector isolated and shut it down in minutes. The government sector never detected it. Detection and isolation, not prevention, decided the outcome.
Nation-State Cyber Threats
The US Treasury sanctioned five Mabna Institute hackers and 30 crypto wallets under Operation Economic Outcast, an expansion of its Iran campaign that also names whole sectors and hands financial-services and crypto compliance teams an urgent screening job.
Phishing
ANY.RUN says the commercial Mirage2FA toolkit hit roughly 4,500 US and EU companies from 2024 to 2026, letting Microsoft 365 users finish a normal login before stealing the authenticated session. It bypasses 2FA, and 48% of targeted addresses were potentially compromised.
Vulnerabilities & Patching
The Shadowserver Foundation has confirmed at least 274 internet-facing Zimbra servers compromised through CVE-2026-73570, the code-injection flaw Synacor patched in ZCS v10.1.20 on July 20. Here is how defenders should triage patching and hunt for signs of compromise.