Amazon has announced that its threat intelligence division has intervened in ongoing cyber operations attributed to hackers associated with Russia’s foreign military intelligence service, the GRU. The activity targeted organizations using Amazon’s cloud infrastructure, with attackers attempting to gain unauthorized access to customer-managed systems.
The company reported that the malicious campaign dates back to 2021 and largely concentrated on Western critical infrastructure. Within this scope, energy-related organizations were among the most frequently targeted sectors, indicating a strategic focus on high-impact industries.
Amazon’s investigation shows that the attackers initially relied on exploiting security weaknesses to break into networks. Over multiple years, they used a combination of newly discovered flaws and already known vulnerabilities in enterprise technologies, including security appliances, collaboration software, and data protection platforms. These weaknesses served as their primary entry points.
As the campaign progressed, the attackers adjusted their approach. By 2025, Amazon observed a reduced reliance on vulnerability exploitation. Instead, the group increasingly targeted customer network edge devices that were incorrectly configured. These included enterprise routers, VPN gateways, network management systems, collaboration tools, and cloud-based project management platforms.
Devices with exposed administrative interfaces or weak security controls became easy targets. By exploiting configuration errors rather than software flaws, the attackers achieved the same long-term goals: maintaining persistent access to critical networks and collecting login credentials for later use.
Amazon noted that this shift reflects a change in operational focus rather than intent. While misconfiguration abuse has been observed since at least 2022, the sustained emphasis on this tactic in 2025 suggests the attackers deliberately scaled back efforts to exploit zero-day and known vulnerabilities. Despite this evolution, their core objectives remained unchanged: credential theft and quiet movement within victim environments using minimal resources and low visibility.
Based on overlapping infrastructure and targeting similarities with previously identified threat groups, Amazon assessed with high confidence that the activity is linked to GRU-associated hackers. The company believes one subgroup, previously identified by external researchers, may be responsible for actions taken after initial compromise as part of a broader, multi-unit campaign.
Although Amazon did not directly observe how data was extracted, forensic evidence suggests passive network monitoring techniques were used. Indicators included delays between initial device compromise and credential usage, as well as unauthorized reuse of legitimate organizational credentials.
The compromised systems were customer-controlled network appliances running on Amazon EC2 instances. Amazon emphasized that no vulnerabilities in AWS services themselves were exploited during these attacks.
Once the activity was detected, Amazon moved to secure affected instances, alerted impacted customers, and shared intelligence with relevant vendors and industry partners. The company stated that coordinated action helped disrupt the attackers’ operations and limit further exposure.
Amazon also released a list of internet addresses linked to the activity but cautioned organizations against blocking them without proper analysis, as they belong to legitimate systems that had been hijacked.
To mitigate similar threats, Amazon recommended immediate steps such as auditing network device configurations, monitoring for credential replay, and closely tracking access to administrative portals. For AWS users, additional measures include isolating management interfaces, tightening security group rules, and enabling monitoring tools like CloudTrail, GuardDuty, and VPC Flow Logs.
The modern internet, though vast and advanced, remains surprisingly delicate. A minor technical fault or human error can disrupt millions of users worldwide, revealing how dependent our lives have become on digital systems.
On October 20, 2025, a technical error in a database service operated by Amazon Web Services (AWS) caused widespread outages across several online platforms. AWS, one of the largest cloud computing providers globally, hosts the infrastructure behind thousands of popular websites and apps. As a result, users found services such as Roblox, Fortnite, Pokémon Go, Snapchat, Slack, and multiple banking platforms temporarily inaccessible. The incident showed how a single malfunction in a key cloud system can paralyze numerous organizations at once.
Such disruptions are not new. In July 2024, a faulty software update from cybersecurity company CrowdStrike crashed around 8.5 million Windows computers globally, producing the infamous “blue screen of death.” Airlines had to cancel tens of thousands of flights, hospitals postponed surgeries, and emergency services across the United States faced interruptions. Businesses reverted to manual operations, with some even switching to cash transactions. The event became a global lesson in how a single rushed software update can cripple essential infrastructure.
History provides many similar warnings. In 1997, a technical glitch at Network Solutions Inc., a major domain registrar, temporarily disabled every website ending in “.com” and “.net.” Though the number of websites was smaller then, the event marked the first large-scale internet failure, showing how dependent the digital world had already become on centralized systems.
Some outages, however, have stemmed from physical damage. In 2011, an elderly woman in Georgia accidentally cut through a fiber-optic cable while scavenging for copper, disconnecting the entire nation of Armenia from the internet. The incident exposed how a single damaged cable could isolate millions of users. Similarly, in 2017, a construction vehicle in South Africa severed a key line, knocking Zimbabwe offline for hours. Even undersea cables face threats, with sharks and other marine life occasionally biting through them, forcing companies like Google to reinforce cables with protective materials.
In 2022, Canada witnessed one of its largest connectivity failures when telecom provider Rogers Communications experienced a system breakdown that halted internet and phone services for roughly a quarter of the country. Emergency calls, hospital appointments, and digital payments were affected nationwide, highlighting the deep societal consequences of a single network failure.
Experts warn that such events will keep occurring. As networks grow more interconnected, even a small mistake or single-point failure can spread rapidly. Cybersecurity analysts emphasize the need for stronger redundancy, slower software rollouts, and diversified cloud dependencies to prevent global disruptions.
The internet connects nearly every part of modern life, yet these incidents remind us that it remains vulnerable. Whether caused by human error, faulty code, or damaged cables, the web’s fragility shows why constant vigilance, better infrastructure planning, and verified information are essential to keeping the world online.
Salesloft, a popular sales engagement platform, has revealed that a breach of its GitHub environment earlier this year played a key role in a recent wave of data theft attacks targeting Salesforce customers.
The company explained that attackers gained access to its GitHub repositories between March and June 2025. During this time, intruders downloaded code, added unauthorized accounts, and created rogue workflows. These actions gave them a foothold that was later used to compromise Drift, Salesloft’s conversational marketing product. Drift integrates with major platforms such as Salesforce and Google Workspace, enabling businesses to automate chat interactions and sales pipelines.
How the breach unfolded
Investigators from cybersecurity firm Mandiant, who were brought in to assist Salesloft, found that the GitHub compromise was the first step in a multi-stage campaign. After the attackers established persistence, they moved into Drift’s cloud infrastructure hosted on Amazon Web Services (AWS). From there, they stole OAuth tokens, digital keys that allow applications to access user accounts without requiring passwords.
These stolen tokens were then exploited in August to infiltrate Salesforce environments belonging to multiple organizations. By abusing the access tokens, attackers were able to view and extract customer support cases. Many of these records contained sensitive information such as cloud service credentials, authentication tokens, and even Snowflake-related access keys.
Impact on organizations
The theft of Salesforce data affected a wide range of technology companies. Attackers specifically sought credentials and secrets that could be reused to gain further access into enterprise systems. According to Salesloft’s August 26 update, the campaign’s primary goal was credential theft rather than direct financial fraud.
Threat intelligence groups have tracked this operation under the identifier UNC6395. Meanwhile, reports also suggest links to known cybercrime groups, although conclusive attribution remains unsettled.
Response and recovery
Salesloft said it has since rotated credentials, hardened its defenses, and isolated Drift’s infrastructure to prevent further abuse. Mandiant confirmed that containment steps have been effective, with no evidence that attackers maintain ongoing access. Current efforts are focused on forensic review and long-term assurance.
Following weeks of precautionary suspensions, Salesloft has now restored its Salesforce integrations. The company has also published detailed instructions to help customers safely resume data synchronization.
The incident underlines the risks of supply-chain style attacks, where a compromise at one service provider can cascade into breaches at many of its customers. It underscores the importance of securing developer accounts, closely monitoring access tokens, and limiting sensitive data shared in support cases.
For organizations, best practices now include regularly rotating OAuth tokens, auditing third-party app permissions, and enforcing stronger segmentation between critical systems.