{
	"id": "78c884cc-c57f-4ae0-9a75-095ddc74d425",
	"created_at": "2026-11-20T02:32:54.64044Z",
	"updated_at": "2026-11-20T16:57:48.795007Z",
	"deleted_at": null,
	"sha1_hash": "708bcce2a732b96e80f421643bc79f41fa94bc69",
	"title": "From Phishing to LLMjacking: Threat Actors Targeting Cloud Billing Accounts for LLM Usage",
	"llm_title": "From Phishing to LLMjacking: Threat Actors Targeting Cloud Billing Accounts for LLM Usage",
	"authors": "Shira Ayal, Andre Maccarone, Wiz",
	"file_creation_date": "0001-01-01T00:00:00Z",
	"file_modification_date": "0001-01-01T00:00:00Z",
	"file_size": 63268,
	"plain_text": "https://www.wiz.io/blog/from-phishing-to-llmjacking\n\nFrom Phishing to LLMjacking: Threat Actors Targeting Cloud\n\nBilling Accounts for LLM Usage\n\nBy Shira Ayal, Andre Maccarone\n\nPublished: 2026-10-20 · Archived: 2026-11-20 02:01:20 UTC\n\nThe Wiz Customer Incident Response Team (CIRT) has investigated a rising threat pattern in cloud environments:\n\nthreat actors hijacking cloud billing accounts to fund unauthorized AI usage. Threat actors are not only using AI as\n\na tool to conduct attacks, they are also compromising victim environments solely to abuse their finances for mass\n\nLarge Language Model (LLM) execution, incurring hundreds of thousands of dollars in AI API charges.\n\nWiz CIRT has investigated multiple intrusions demonstrating this pattern, with this blogpost focusing specifically\n\non campaigns targeting Google Cloud Platform (GCP) infrastructure. In these attacks, threat actors compromised\n\nprivileged GCP user credentials and linked external, attacker-controlled GCP projects directly to the victim's\n\ninternal billing accounts. Because the AI workloads were processed in foreign projects, prompt contents and\n\nmodel outputs remained completely hidden from the victim organization, leaving the nature of the mass AI usage\n\nunknown.\n\nThis blogpost details the full attack lifecycle observed across these intrusions, from the initial phishing and policy\n\nmanipulation to cross-tenant project linking, and provides actionable threat hunting guidance to help organizations\n\ndetect LLMjacking in their environments.\n\nInitial Access: Phishing via Sponsored Search Ads\n\nThe attack began with a classic yet efficient phishing tactic. The threat actor purchased Google Search ads for\n\ncommon administrative search terms such as “google cloud” and “gcp”. When a user clicked the malicious ad,\n\nthey were led through a Google Sites redirect to a fake Google login web page. Victim users with valid access to\n\nGoogle Console proceeded to input their account credentials and multi-factor-authentication (MFA) codes into the\n\nphishing page. The malicious site captured the user supplied credentials before seamlessly redirecting the user\n\nback to the legitimate Google authentication prompt. Forced to authenticate a second time, the user likely assumed\n\nthe initial attempt was a minor glitch. In reality, the automated phishing kit immediately leveraged the harvested\n\ncredentials to execute the first malicious login into the victim’s GCP environment from  212.73.134[.]136 , the\n\nsource IP address of the phishing kit. The threat actor then reauthenticated a few minutes later from a second IP\n\naddress, which was later used for the rest of the GCP activity. Following reports of this activity, Google Ads has\n\ntaken action to remove the malicious ad campaign.\n\nGoogle automatically flagged the anomalous login attempt as suspicious and dispatched security alert emails to\n\nthe user's inbox. However, to catch such intrusions early, organizations should not rely on user awareness alone,\n\nbut ingest and monitor Google Workspace User Log Events for suspicious login flags and anomalous re-\n\nauthentication triggers.\n\nPage 1 of 5\n\nhttps://www.wiz.io/blog/from-phishing-to-llmjacking\n\nPersistence and Privilege Escalation: Disabling Guardrails and Externalizing\n\nBilling\n\nOnce inside the environment, the threat actor leveraged the compromised account’s excessive privileges to\n\nsolidify their access, bypass organizational security controls, and prepare the billing architecture for external\n\nexploitation.\n\nEnabling APIs and Minting Service Account Credentials\n\nRather than creating new infrastructure from scratch, the attacker targeted existing, low-usage GCP projects to:\n\nEnable the Gemini API product\n\nCreate service accounts with Gemini API keys\n\nAdd the newly created service accounts to high privileged roles such as  roles/ml.admin  and\n\nroles/mcp.toolUser\n\nDismantling Organizational Policy Controls\n\nTo expand their reach without raising immediate alarms within internal domain governance, the threat actor\n\ndeleted safety policies at the organization level, such as  iam.allowedPolicyMemberDomains . This action removed\n\nthe guardrail that restricts Identity and Access Management (IAM) membership to verified company domains,\n\nenabling external, personal accounts to be granted access as well as high-privilege permissions.\n\nInviting External Threat Actor Accounts and Elevating Privileges\n\nWith domain restrictions lifted, the compromised user account was leveraged to add external actor-controlled\n\nGmail accounts directly into the organization's billing account.\n\nThe threat actor also assigned administrative roles to the compromised users or the external accounts, such as:\n\nroles/recommender.billingAccountCudAdmin  on GCP projects\n\nroles/billing.admin  on GCP billing accounts\n\nThe role  roles/billing.admin  granted full control over the billing account, including the ability to manage\n\npermissions, attach external projects, and alter payment settings.\n\nLinking Foreign Projects for Mass Exploitation\n\nThe threat actor linked external GCP projects directly to the victim’s internal billing account. This move\n\nestablished a cross-tenant access, allowing all API requests executed in the threat actors external projects to be\n\nautomatically charged to the victim's billing account while keeping prompt payloads invisible to internal log\n\nmonitoring.\n\nDefence Evasion: Geographic Blending and Admin Lockout\n\nPage 2 of 5\n\nhttps://www.wiz.io/blog/from-phishing-to-llmjacking\n\nOnce initial access was secured, threat actors employed defense evasion techniques designed to avoid triggering\n\nlocation-based security alerts and prevent security teams from regaining control.\n\nIn each of the intrusions investigated, Wiz CIRT observed that after the initial login, subsequent threat actor\n\nactivity shifted to originate from IP addresses geolocated within each victim organization's headquartered country,\n\neven when this did not match the compromised user’s actual physical location.\n\nAdditionally, after securing administrative control, the threat actors demoted pre-existing internal billing\n\nadministrators from their administrative roles down to read-only billing viewers. This modification stripped\n\nlegitimate administrators of their ability to revoke permissions or detach rogue projects directly from the console.\n\nImpact: AI Resource Abuse\n\nThe primary objective of these intrusions was non-destructive financial exploitation - leveraging hijacked cloud\n\ninfrastructure to offload the costs of high-volume LLM inference. Across the observed cases, the impact\n\nmanifested as rapid billing surges driven by automated, large-scale Gemini API utilization.\n\nWithin 24 to 48 hours of initial access, victim organizations experienced unexpected charges reaching tens to\n\nhundreds of thousands of dollars in unauthorized Gemini API usage. This activity represented a sharp surge from\n\nthe organizations' established baseline compute spend.\n\nBecause the threat actors linked external, attacker-controlled GCP projects directly to the victim’s billing accounts,\n\nall API key generation, prompt requests, and model outputs occurred entirely within foreign tenants.\n\nConsequently, while the victim organization carried financial responsibility for the API consumption, internal\n\nsecurity and audit logs provided no visibility into the nature, content, or context of the AI workloads being\n\nexecuted.\n\nDetection and Threat Hunting Indicators\n\nTo detect potential LLMjacking and unauthorized billing account manipulation in your Google Cloud\n\nenvironment, security teams should hunt for indicators across Google Workspace, GCP Admin Activity, and Cloud\n\nBilling audit logs. Wiz customers can review the Threat Intel Center advisory that includes the following\n\nindicators:\n\nReview Google Workspace and GCP audit logs for activity from the indicators of compromise section\n\nbelow. Wiz customers can use this Cloud Events query.\n\nReview GCP IAM for any users with external domains, especially non-corporate email providers such as\n\n@gmail.com. Wiz customers can use this Cloud Events query, and replace the domain.com with their own\n\norganization domain.\n\nReview GCP audit logs for changes to organization policy controls, especially\n\nallowedPolicyMemberDomains , and IAM policy changes related to billing account administrators. Wiz\n\ncustomers can use this Cloud Events query.\n\nPage 3 of 5\n\nhttps://www.wiz.io/blog/from-phishing-to-llmjacking\n\nReview GCP audit logs for unexpected Gemini API enablement, service account creation and service\n\naccount API key creation in old projects Wiz customers can use this Cloud Events query.\n\nMonitor abnormal cost surge in API usage, and potentially enforce spend cap budgets. Wiz Cloud Cost\n\ncustomers can detect abnormal AI usage per model and cloud provider by monitoring Cost Monitor\n\nFindings for Resource Anomalies.\n\nWiz Defend customers ingesting cloud logs such as Google Workspace will get Wiz Threats on the initial logins.\n\nCustomers ingesting GCP audit logs will get Threats for the described activity.\n\nIf you experience a breach or any cloud security incident, contact Wiz CIRT for incident response support.\n\nIndicators of Compromise\n\nIOC Type Description\n\n212.73.134[.]136 IP Address Attacker infrastructure - IP seen in the initial login from the phishing page\n\n*.gboogl[.]com Domain Attacker infrastructure - phishing domain\n\n*.groosl[.]com Domain Attacker infrastructure - phishing domain\n\n*.gprokl[.]com Domain Attacker infrastructure - phishing domain\n\n*.gxoosl[.]com Domain Attacker infrastructure - phishing domain\n\n*.gsoolg[.]com Domain Attacker infrastructure - phishing domain\n\n*.gwoogl[.]com Domain Attacker infrastructure - phishing domain\n\n*gzoozl[.]com Domain Attacker infrastructure - phishing domain\n\n*groonl[.]com Domain Attacker infrastructure - phishing domain\n\n*gsool[.]space Domain Attacker infrastructure - phishing domain\n\n*gbrokl[.]space Domain Attacker infrastructure - phishing domain\n\n*gbrokl[.]xyz Domain Attacker infrastructure - phishing domain\n\n*gsoosl[.]xyz Domain Attacker infrastructure - phishing domain\n\n*gsool[.]xyz Domain Attacker infrastructure - phishing domain\n\n*gpoosl[.]com Domain Attacker infrastructure - phishing domain\n\n*groolg[.]com Domain Attacker infrastructure - phishing domain\n\n*gcslom[.]com Domain Attacker infrastructure - phishing domain\n\n*gcoalg[.]com Domain Attacker infrastructure - phishing domain\n\nPage 4 of 5\n\nhttps://www.wiz.io/blog/from-phishing-to-llmjacking\n\nSource: https://www.wiz.io/blog/from-phishing-to-llmjacking\n\nPage 5 of 5",
	"extraction_quality": 1,
	"language": "EN",
	"sources": [
		"OTX"
	],
	"origins": [
		"web"
	],
	"references": [
		"https://www.wiz.io/blog/from-phishing-to-llmjacking"
	],
	"report_names": [
		"from-phishing-to-llmjacking"
	],
	"threat_actors": [],
	"ts_created_at": 1795141974,
	"ts_updated_at": 1795193868,
	"ts_creation_date": 0,
	"ts_modification_date": 0,
	"files": {
		"pdf": "https://archive.orkl.eu/708bcce2a732b96e80f421643bc79f41fa94bc69.pdf",
		"text": "https://archive.orkl.eu/708bcce2a732b96e80f421643bc79f41fa94bc69.txt",
		"img": "https://archive.orkl.eu/708bcce2a732b96e80f421643bc79f41fa94bc69.jpg"
	},
	"iocs": [
		{
			"type": "domain",
			"value": "domain.com",
			"count": 1,
			"flags": [],
			"some_flags": []
		},
		{
			"type": "domain",
			"value": "gboogl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gbrokl.space",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gbrokl.xyz",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gcoalg.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gcslom.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gmail.com",
			"count": 1,
			"flags": [],
			"some_flags": []
		},
		{
			"type": "domain",
			"value": "gpoosl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gprokl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "groolg.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "groonl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "groosl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gsoolg.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gsool.space",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gsool.xyz",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gsoosl.xyz",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gwoogl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gxoosl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "gzoozl.com",
			"count": 1,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "domain",
			"value": "www.wiz.io",
			"count": 6,
			"flags": [
				"in_url_host"
			],
			"some_flags": [
				"in_url_host"
			]
		},
		{
			"type": "ipv4",
			"value": "212.73.134.136",
			"count": 2,
			"flags": [
				"defanged"
			],
			"some_flags": [
				"defanged"
			]
		},
		{
			"type": "url",
			"value": "https://www.wiz.io/blog/from-phishing-to-llmjacking",
			"count": 6,
			"flags": [],
			"some_flags": []
		}
	]
}