Amgen Cloud Data Breach Highlights the Risks of Sensitive Information Stored Across Third-Party Environments

The data breach disclosed by biotechnology company Amgen demonstrates how an organization’s cybersecurity risk now extends far beyond the systems and infrastructure it directly operates. Amgen reported that threat actors obtained company information and patient data from multiple cloud environments operated by third-party service providers. The information confirmed as having been exfiltrated includes proprietary company data, patient protected health information and other information, while the company is continuing to determine whether the attackers also accessed or stole additional patient information, confidential business records, intellectual property, research and development material or other sensitive information.

The incident is particularly significant because it potentially combines two highly valuable categories of data: sensitive patient information and proprietary pharmaceutical research. Patient information can support identity theft, fraud, extortion and targeted social-engineering attacks, while intellectual property and research information may have commercial, competitive or strategic value. A breach involving both categories can therefore create risks for patients, business operations, regulatory compliance, corporate reputation and future research programmes.

The breach was discovered in July 2026

Amgen detected unauthorised activity during July 2026, activated its cybersecurity incident-response plan, implemented containment measures and engaged independent forensic specialists to investigate the incident. On July 29, the company determined that the breach was material after considering the apparent volume of affected files and the possibility that the information within those files was sensitive.

The company has not yet publicly disclosed the identity of the affected cloud service providers, the precise date on which the intrusion began, how the attackers initially gained access or how long they remained inside the affected environments. It has also not disclosed the number of patients or other individuals affected, the exact categories of patient information exposed, whether compromised credentials were involved, whether the incident included ransomware or data extortion, or whether a known threat group was responsible.

These unanswered questions are important because the method of compromise will determine many of the broader lessons from the incident. An exposed cloud-storage service, stolen administrative credentials, compromised single sign-on account, malicious third-party employee, improperly configured application or exploited cloud integration can each require a very different security response.

Cloud security is a shared responsibility

Organizations frequently move sensitive information to cloud platforms because cloud services offer scalability, availability, flexibility and reduced infrastructure-management requirements. However, moving information to the cloud does not transfer the organization’s responsibility for protecting it, because cloud security operates under a shared-responsibility model.

The cloud provider may secure the underlying physical infrastructure, hypervisor, storage platform and core service, while the customer must still protect identities, permissions, data, applications, integrations, access tokens, encryption keys and administrative configurations. In software-as-a-service environments, customers may have less responsibility for the underlying platform, but they remain responsible for determining who is authorised to access the service, what information is uploaded to it, whether permissions are excessive, how administrator accounts are protected, whether third-party integrations are secure, how data is classified and monitored, whether audit logs are available and reviewed, and how quickly access can be revoked after an incident.

The phrase “third-party cloud environment” should therefore never be interpreted as “someone else’s security problem.” If an organization’s sensitive information is stored there, the resulting breach remains the organization’s operational, regulatory and reputational problem. Outsourcing infrastructure may offer substantial benefits, but outsourcing accountability remains a fantasy repeatedly disproved by breach notifications.

Multiple cloud environments increase complexity

Amgen’s disclosure refers to data being exfiltrated from multiple cloud environments operated by third-party providers, which highlights the security challenges created by modern multi-cloud and software-as-a-service environments. Large organizations rarely use a single cloud platform and may maintain information across public cloud infrastructure, managed storage services, collaboration platforms, research applications, patient-support systems, customer relationship management platforms, file-sharing services, data analytics environments, external research portals, third-party business applications, and development and testing systems.

Each environment may use different security controls, authentication mechanisms, administrative interfaces, logging formats and retention policies. Security teams must therefore maintain visibility across systems they do not fully control, and when controls are fragmented, attackers can exploit the gaps between them.

For example, a security team may closely monitor its primary corporate network while having limited visibility into downloads from a third-party application. An attacker using a legitimate account may therefore access and extract data without generating a traditional malware alert inside the organization’s own infrastructure.

Identity is becoming the new security perimeter

Although the exact intrusion method has not been disclosed, cloud breaches frequently involve compromised credentials, stolen session tokens, weak multi-factor authentication, excessive permissions or abuse of trusted third-party accounts. Traditional network boundaries provide less protection when sensitive applications are accessible through the internet and employees, vendors and automated systems can connect from multiple locations.

In this environment, identity becomes one of the most important security controls. Organizations should secure cloud access using phishing-resistant multi-factor authentication, conditional access policies, least-privilege permissions, privileged access management, short-lived access tokens and strong session controls. They should also conduct regular access reviews, immediately remove dormant accounts, use separate administrative identities, restrict third-party application authorisations and monitor unusual login behaviour.

Authentication alone should not be treated as proof that an activity is legitimate, because once an attacker obtains valid credentials, every subsequent request may appear to come from an authorised user. Security systems must therefore evaluate what the authenticated identity is doing. A valid user downloading an unusual volume of patient records, accessing information outside normal working hours, connecting from an unfamiliar device or exporting data that the user has never previously accessed should receive additional scrutiny.

Cloud audit logs must be actively monitored

Cloud providers and software-as-a-service platforms often generate detailed audit records that may show login attempts, successful authentications, administrative changes, file access, file downloads, permission modifications, creation of new access tokens, addition of third-party applications, data exports, changes to retention policies, creation of forwarding or sharing rules, and access from new devices or locations. However, these logs provide little protection when they are not enabled, retained or analysed.

Organizations should ensure that security logs from every critical cloud service are collected centrally and correlated with endpoint, identity and network activity. This correlation is particularly valuable during an investigation because a cloud platform may show that a user downloaded thousands of files, identity logs may show that the user authenticated from an unusual source, endpoint records may show no corresponding legitimate activity from the employee’s managed device, and network information may show that the transfer bypassed the corporate environment entirely. Individually, each event may appear incomplete, but together they can reveal account compromise and data exfiltration.

Sensitive data must be classified before it can be protected

The Amgen incident reportedly involved patient protected health information and proprietary company information, while the investigation is examining whether intellectual property and research and development data may also have been affected. These different categories of information require different protection, retention and access controls.

Organizations cannot protect sensitive information effectively unless they know what data they hold, where it is stored, why it is retained, who can access it, which applications process it, which third parties receive it, how it moves between environments and how long it should remain available. Data classification should therefore be implemented across on-premises, cloud and third-party systems.

Patient information, research information, intellectual property, financial data and ordinary operational records should not all receive identical treatment. Highly sensitive information may require stronger encryption, more restrictive access permissions, additional monitoring, shorter retention periods and tighter controls over copying, downloading and sharing.

Data inventory is equally important because organizations often know where primary databases are located but have far less visibility into duplicate copies created through exports, email attachments, collaboration platforms, test environments, backup systems and third-party workflows. Attackers do not care whether the stolen file was the authorised master record or an inconveniently forgotten export created six months earlier. Sensitive information remains sensitive wherever bureaucracy has duplicated it.

Healthcare information has long-term value

Protected health information can remain valuable to criminals for many years because a payment-card number can be cancelled and a password can be reset, while medical history, date of birth and other identity information cannot be replaced so easily. Depending on the information involved, healthcare data can be used for identity theft, insurance fraud, prescription fraud, financial fraud, tax fraud, account recovery attacks, blackmail, extortion, targeted phishing, impersonation of healthcare providers and social engineering against patients and family members.

Attackers may also combine data from several breaches to create a more complete profile of an individual. A future phishing message may refer to a real medical condition, treatment programme, medication or healthcare provider, making the communication appear far more credible and increasing the likelihood that the recipient will respond.

Organizations handling patient information must therefore consider not merely the immediate impact of the breach but also the potential for long-term misuse.

Proprietary and research data introduce additional risks

The potential exposure of confidential business information, intellectual property and research and development material creates a different set of risks. Biotechnology and pharmaceutical organizations invest significant time and resources in drug discovery, clinical research, manufacturing methods and regulatory submissions.

Stolen research information may potentially be used for corporate espionage, competitive intelligence, extortion, manipulation of commercial negotiations, targeting of research partners, attacks against clinical trial participants, fraud involving unpublished research and compromise of additional systems through exposed technical information.

Research environments are particularly difficult to secure because they often depend on collaboration between employees, laboratories, universities, healthcare providers, contract research organizations and external specialists. Security controls must therefore protect information without making legitimate scientific collaboration impossible, which requires granular access control rather than broad permissions granted merely for convenience.

Third-party risk assessments must extend beyond questionnaires

The breach also raises questions about how organizations assess and continuously monitor third-party cloud providers. Vendor security assessments frequently depend on annual questionnaires, compliance certificates and contractual assurances, but these measures provide only a snapshot of risk at a particular point in time.

A provider may pass a security review and later experience credential compromise, configuration errors, vulnerabilities, changes in subcontractors, excessive administrator access, inadequate logging, poor incident detection, failure to remove former employees, unauthorised application integrations or delayed breach notification. Organizations should therefore evaluate third-party risk throughout the relationship rather than only during procurement.

Contracts should clearly establish requirements for security controls, encryption, access management, audit logging, incident notification, cooperation during forensic investigations, data retention, data deletion, subcontractor management, regulatory compliance, independent security testing and evidence preservation. Organizations should also maintain the practical ability to obtain required logs and investigate incidents, because a contractual right to audit is far less reassuring when exercising it requires several committees, multiple legal reviews and the geological patience normally associated with continental drift.

Data exfiltration may not disrupt operations

Amgen has stated that the incident has not, to date, affected its products, manufacturing operations, financial reporting systems or ability to meet patient needs. While this is important because the event does not currently appear to have caused an operational shutdown, the absence of disruption should not reduce the seriousness of a confidentiality breach.

Modern attackers increasingly focus on stealing information rather than encrypting systems. An organization may continue operating normally while data is quietly copied from cloud repositories, creating a dangerous visibility problem.

Traditional incident detection often focuses on malware execution, system crashes, encryption activity, service interruption, unusual processor usage and network outages. Pure data-theft attacks may produce none of these obvious symptoms because the attacker may use valid credentials, legitimate application programming interfaces and standard file-download functionality.

Organizations therefore need controls capable of identifying abnormal data access even when the underlying platform is operating exactly as designed.

Data loss prevention must cover cloud applications

Data loss prevention should not be limited to endpoint devices or outbound email. Modern DLP controls must account for cloud storage, collaboration tools, web applications, application programming interfaces and unmanaged devices.

Useful controls may include detecting bulk downloads, restricting downloads to unmanaged devices, blocking public sharing of sensitive files, applying sensitivity labels, monitoring unusual file access, preventing unauthorised data exports, controlling copying between cloud services, restricting the use of personal cloud-storage accounts, inspecting uploads to unapproved platforms and alerting when confidential information is accessed by unusual identities.

These controls should use business context. A researcher downloading files required for an approved project may be legitimate, while an administrative account downloading the same research repository at an unusual hour from an unknown device may require immediate investigation.

Incident response must include third-party environments

Organizations should update their incident-response plans to account for breaches occurring outside their directly managed network. A cloud-focused response plan should establish who contacts the service provider, how logs will be obtained, how compromised accounts will be disabled, how access tokens will be revoked, how active sessions will be terminated, how encryption keys and credentials will be rotated, how evidence will be preserved, how affected information will be identified, how legal and regulatory obligations will be assessed, how patients and other individuals will be notified, and how business operations will continue during containment.

Speed matters because if an organization must first determine who owns the vendor relationship or negotiate access to logs after discovering a breach, valuable evidence may disappear while the attacker retains access.

Network security remains relevant in cloud breaches

Cloud applications may operate outside the traditional enterprise network, but network security still plays an important role. Organizations can use network and application-level visibility to identify access to unusual cloud services, connections from unauthorised applications, large data transfers, communication with newly observed destinations, use of anonymisation services, access from unmanaged devices, abnormal upload and download patterns and attempts to bypass approved cloud gateways.

However, destination-based controls alone are insufficient because a connection to a legitimate cloud provider cannot automatically be considered safe. The security system must understand the user, application, device, data and purpose associated with the connection. Contextual visibility helps distinguish ordinary cloud usage from suspicious activity occurring through a legitimate service.

The larger cybersecurity lesson

The Amgen breach demonstrates that an organization can have secure manufacturing systems, functioning products and uninterrupted business operations while still experiencing a serious data-security incident. Cybersecurity must therefore protect confidentiality as well as availability, and organizations need to move beyond asking whether systems are running to asking whether information is being accessed appropriately.

Effective protection requires a coordinated security architecture that includes strong identity controls, cloud security monitoring, data classification, least-privilege access, context-aware data loss prevention, centralised logging, third-party risk management, network segmentation, behavioural analytics, tested incident-response procedures and continuous review of cloud permissions and integrations.

The incident also demonstrates why third-party cloud services must be treated as part of the organization’s own attack surface. The infrastructure may belong to another company, but the information, regulatory obligations and consequences belong to the organization that entrusted its data to that environment.

As investigations continue, more details may emerge regarding the attack method, affected cloud platforms and number of patients involved. The existing disclosure already provides an important warning: sensitive information distributed across external cloud environments can be difficult to inventory, monitor and protect unless security controls follow the data wherever it is stored or processed.

Organizations should not wait for a breach to discover where their most sensitive information has travelled. They need visibility before the attacker develops an inventory on their behalf.

Pharmaceutical company Amgen says it suffered a data breach after threat actors stole corporate data and patient information stored in multiple cloud systems operated by third-party service providers. [...]

Source: Amgen says cloud data breach exposed patient health, proprietary info via Bleeping Computer — published 31 Jul 2026.