The breach wasn’t a police system — it was the vendor holding 135,000 identities. Extortion group ExfilSquad claims it exfiltrated roughly 1.9GB / ~135,000 records from the UK’s Police National Legal Database (PNLD) — its own figures put it at ~114,000 PNLD subscribers plus ~21,000 users of the public “Ask the Police” service — and has published a sample on the dark web with a demand to pay or see the rest released. The exposed material is contact-tier data — names, organizations, and work email addresses of police officers, staff, criminal-justice professionals, and government partners — not case files, but for a law-enforcement population that distinction still matters.
UK authorities are investigating; ExfilSquad is the same actor that recently claimed a UK Department for Education breach, and its pitch (“the payment we request is simply a rounding error compared to the litigation costs of your data leaking”) is the extortion economics of 2026 stated out loud. The durable read isn’t “patch PNLD” — it’s that your workforce’s identity data increasingly lives concentrated inside third-party SaaS and reference databases you don’t run and can’t see into, and that concentration is the target. For every outside system that holds your people’s identities: can you name it, do you know what it holds, and would you even know if it left?
Sources: The Hacker News · Computing · TechRadar
Cyberattacks & Breaches
PNLD extortion breach exposes ~135,000 UK police and government contacts. ExfilSquad dumped a sample and is running a pay-or-leak play against a third-party legal database, not a police network. (The Hacker News; Computing; TechRadar)
Adform’s ad-tech script was poisoned to steal crypto. Attackers compromised Adform’s widely embedded trackpoint-async.js tracking script so it watched visitors’ clipboards for Bitcoin, Ethereum, and TRON wallet addresses and silently swapped in attacker addresses across downstream sites. Researcher Kevin Beaumont surfaced it; Adform says first malicious activity traces to July 26, detection came July 27, the code has been removed, clients notified, and authorities informed — and it’s telling users to clear browser cache since the tampered file may still be cached. Adform serves roughly 14,000 businesses and holds a large share of the demand-side ad market, so a single trusted script becomes a supply-chain delivery channel. (BleepingComputer; The Hacker News; DoublePulsar; Adform)
Framework customer data leaked via an upstream zero-day at its database provider. Framework (the modular-laptop maker) disclosed that customer names, email addresses, phone numbers, and physical addresses were exposed through a zero-day attack against a third-party database provider (Metabase), not a compromise of Framework’s own systems. It’s a small breach in raw numbers but a clean example of the pattern: your customer data is only as contained as the least-secure vendor in your stack.
So what: Three unrelated incidents, one shape — the loss happened in someone else’s system that your people or customers trusted. Inventory the third parties that hold your identity and customer data, demand breach-notification and telemetry commitments in writing, and treat “we don’t run it” as a reason to watch it harder, not less.
Vulnerabilities & Exposure
JetBrains TeamCity unauthenticated RCE (CVE-2026-63077) is being exploited — CISA KEV, deadline already passed. CISA added CVE-2026-63077 to the Known Exploited Vulnerabilities catalog on August 5 with an unusually short remediation deadline of August 8 (now behind us). The flaw is a CWE-502 deserialization bug in TeamCity’s agent-polling protocol that lets an unauthenticated attacker with network access run OS commands as the TeamCity server process; it affects all On-Premises releases before 2025.11.7 and 2026.1.3, and JetBrains patched it in late July. Because TeamCity sits at the center of CI/CD, a takeover means stealing stored credentials, tampering with build outputs, and inserting code into everything downstream. (CISA; Rapid7; SecurityWeek; The Hacker News)
So what: If you run TeamCity On-Prem and it’s reachable, you are past the deadline and inside the exploitation window — patch to 2025.11.7 / 2026.1.3 now, then assume-compromise: rotate CI/CD secrets, review recent build artifacts for tampering, and check for unexpected server-side command execution. A patched build server that already shipped a poisoned artifact isn’t actually fixed.
AI Impact on Security
The software delivery pipeline is becoming the AI-era attack surface. Line up this weekend’s TeamCity CI/CD RCE and the Adform trusted-script poisoning next to last week’s Langflow (agentic-AI platform) RCE and the RovoBlast assistant exfiltration, and the pattern is hard to miss: as more code and action flows through automated pipelines and sanctioned tools — increasingly authored or triggered by AI coding assistants and agents — the pipeline’s trust boundary is where compromise now pays off. The failure mode isn’t a jailbroken model; it’s that build systems, agent platforms, and embedded scripts trust inputs and callers they were never scoped to trust, and every downstream consumer inherits that trust.
So what: The defensive move is governing the harness, not buying a smarter model — treat CI/CD, AI-workflow platforms, and every deployed assistant as production identities with least-privilege scoping, provenance on what they produce, and telemetry that can tell “normal automation” from “automation acting on a hostile instruction.” This is a visibility-and-data-foundation problem before it’s an AI problem.
Regulatory & Policy
EU AI Act’s August 2 milestone passed — transparency obligations live, high-risk deadline in limbo. As of August 2, 2026, the AI Act’s Article 50 transparency obligations (labeling AI interactions and AI-generated content) are in force. The higher-stakes Annex III high-risk-system obligations were also due August 2, but a proposed “Digital Omnibus” simplification would defer them to December 2027; until that text is formally published in the EU Official Journal, the original August 2 deadline legally stands as written. Net: transparency duties apply now, high-risk timing is genuinely uncertain, and “we assumed it was delayed” is not a compliance posture. (Travers Smith; DLA Piper)
So what: If you build or deploy AI in the EU, confirm your transparency/labeling controls are actually live today, and keep your high-risk conformity work moving rather than betting on the deferral — the safe read is “comply on the original timeline until the Official Journal says otherwise.”
Sources & further reading (14)
- The Hacker News — “PNLD Breach Exposes U.K. Police and Government Contact Details on Dark Web”
- Computing — “UK police legal database breach exposes details of more than 100,000 officers and staff”
- TechRadar — “Over 100,000 UK Police and staff have personal data leaked in attack on national database”
- BleepingComputer — “Online ad firm Adform’s script compromised to steal cryptocurrency”
- The Hacker News — “Hackers Poison Adform Script to Swap Crypto Wallet Addresses Across Customer Sites”
- DoublePulsar (Kevin Beaumont) — “Adform compromised to serve crypto stealer via supply chain attack”
- Adform — “Security Incident — Company Update”
- this week in security — “August 9 2026 edition” (Framework / Metabase)
- CISA — “CISA Adds One Known Exploited Vulnerability to Catalog” (TeamCity CVE-2026-63077)
- Rapid7 — “Analysis of CVE-2026-63077, an unauthenticated RCE in JetBrains TeamCity”
- SecurityWeek — “Hackers Start Exploiting Recent JetBrains TeamCity Vulnerability”
- The Hacker News — “CISA Flags TeamCity CVE-2026-63077 RCE Flaw Under Active Exploitation”
- Travers Smith — “EU agrees to delay key AI Act compliance deadlines”
- DLA Piper — “The Digital AI Omnibus: Proposed deferral of high-risk AI obligations under the AI Act”