The Tech Leaders Brief

Mandiant reports LLM theft generating daily charges above US$100,000

3 selected stories, with article evidence, consequences and limits.

About 3 minutes.

Research coverage: 6 of 6 candidate articles read. Analysis is model-generated; source links show its evidence.

1 linked primary-source reads failed; unavailable text was excluded from analysis.

Independent

Mandiant reports multiple 2026 incidents involving stolen enterprise model access, with fraudulent consumption charges exceeding US$100,000 in a day and cumulative charges reaching US$1 million at some organizations before access was revoked.

Why it matters: Cloud owners face this exposure even without previously enabled LLM services: Mandiant observed attackers enabling services after compromise. In another incident, an attacker created and deleted short-lived service accounts to bypass rate limits. Permissions to enable services and create identities can therefore expand the spending accessible through a compromised credential.

Decision: Where cloud identities can enable model services, restrict that permission to roles that require it and monitor unauthorized enablement, including in projects that do not currently use LLMs.

What remains uncertain: These are vendor-reported incident observations. They do not establish prevalence or how much fraudulent billing victims ultimately paid. The reported service-account rotation does not establish that every type of quota or spending control was bypassed.

Read: Reported article: Token thieves costing companies $100K+ per day in illicit LLM charges; Primary source: The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources | Google Cloud Blog

ARTEX campaigns widen targeting and accelerate cloud access

Independent

CrowdStrike says VAPOR PANDA used ARTEX to target over 200 organizations in under a fortnight, with confirmed post-exploitation activity at nine. Separately, Microsoft reports Storm-0940 attacks that reached privileged cloud access in less than three hours in multiple instances.

Why it matters: Microsoft observed attackers using compromised ConnectWise application credentials and existing tenant permissions without interactive user sign-in. Subsequent application credential changes could preserve access after remediation of the original application. For affected Microsoft 365 operators, disabling the vendor connection may therefore leave additional unauthorized access intact.

Decision: For tenants where affected ConnectWise applications were authorized, verify that containment also addresses subsequent unauthorized credentials and permissions. Preserve evidence before removing confirmed unauthorized access.

What remains uncertain: CrowdStrike's primary report could not be read; its findings come through iTnews. Microsoft's separate campaign report does not independently confirm them. Microsoft says most targeting failed and authentication alone does not establish data theft. Reported attempts to modify privileges should not all be counted as successful escalation. GTIG's quoted assessment found no evidence of fully human-free hacking.

Read: Reported article: ARTEX AI hacking framework spreads to another China-linked group - iTnews; Primary source: AI-driven, large-scale nation-state campaign using ARTEX enables supply chain compromise | Microsoft Security Blog

Westpac forecasts warehouse migration cost at half its original budget

Independent

Westpac says its St.George warehouse migration will cost around $35 million against a $70 million budget. CEO Anthony Miller says AI tools accelerated code conversion, testing and documentation, helping complete the migration more than two years early.

Why it matters: The bank attributes shorter delivery and lower project costs to AI assistance within migration tasks. For transformation leaders, this provides a concrete case to evaluate against their own migration workloads. Westpac separately expects around $15 million in annual direct benefits, with benefits beginning in FY27 and legacy warehouse decommissioning planned for that year.

Decision: For comparable migrations, evaluate AI assistance against the project's cost and delivery baseline; do not use Westpac's reported 50 percent reduction as an assumed saving.

What remains uncertain: The figures are Westpac's claims reported by iTnews; the referenced investor documents were not supplied. Final cost and recurring benefits remain prospective. The evidence neither isolates AI's contribution from other delivery factors nor demonstrates repeatability. It does not establish that all benefits depend on completed decommissioning.

Read: Reported article: Westpac reveals AI cut data warehouse migration cost from $70m to $35m - iTnews

No linked primary source could be read; claims rely on the reporting above.

Sources

Source types: Vendor Independent Regulator "Vendor claim, unconfirmed" means no independent or regulator item in this edition corroborates it.

Independent outlets scanned: iTnews, The Register, Ars Technica, BleepingComputer, Dark Reading, MIT Technology Review. Regulator source: CISA (US only). Not monitored: Cyber.gov.au (no reachable official feed), so there is no Australian regulator coverage.