AWS Unveils Significant Advancements at Summit: Introducing S3 Files and Enhanced Security Tools
At its recent Summit, AWS announced two major developments that extend beyond typical product updates. The new S3 Files feature expands object storage to include network file system access, eliminating the need for data migration. In cybersecurity, AWS introduced the Claude Mythos Preview via Amazon Bedrock and announced the general availability of the AWS Security Agent, an innovative tool that actively tests identified vulnerabilities before alerting users.
For years, object storage and file storage have existed in separate architectures, each optimized for specific use cases. Amazon S3 has established itself as the benchmark for managing large volumes of data, powering analytic pipelines and cloud-native workloads. However, many applications, data processing tools, machine learning pipelines, and Unix utilities still interact with data through file-oriented POSIX interfaces. Until now, IT teams had to maintain synchronized copies between S3 and mounted file systems or modify their applications to directly utilize the S3 API. S3 Files aims to eliminate these friction points.
This new functionality integrates Amazon EFS with S3, allowing users to mount any bucket or prefix as a network file system in an EC2 instance, container, or Lambda function. Standard libraries like pandas, machine learning pipelines, and Unix utilities can now interact with S3 data without necessitating any code changes or data relocation. In the event of a conflict between modifications made via file and object interfaces, S3 serves as the single source of truth, overriding any alternate versions.
Gradual Import, Synchronization, and Read Performance
The S3 Files loading model is designed to accommodate buckets containing millions of objects without an initial delay. When connected, files smaller than 128 KB are imported immediately into the file system. For larger files, only metadata is loaded, while the content is retrieved at the time of access. This approach prevents bandwidth congestion during initialization while ensuring immediate availability of the complete namespace. The synchronization is bidirectional and automatic, with any changes made via the file interface propagated back to S3 roughly every sixty seconds. For intensive sequential reads, a bypass function redirects requests directly to S3 using parallel GET calls, boasting a reported throughput of 3 GB/s per client.
While S3 Files enhances access efficiency, it does not alter AWS’s existing egress transfer pricing model. Each file read triggers a GET request to S3, and reads within the same region incur no egress charges; however, any read from a different region or external to the AWS network will be billed at standard egress rates. Additionally, the bidirectional synchronization generates a volume of PUT, GET, and LIST API calls that can accumulate on highly active buckets. CIOs who maintained architectures with local copies to avoid repeated S3 calls must assess the financial implications before transitioning to S3 Files. The promised simplification of architectural complexity does not automatically equate to lower operating costs.
S3 Files Completes a Trifecta of Multimodal Storage Solutions
The introduction of S3 Files marks a continuation of a trend begun at re:Invent 2024, which saw the launch of S3 Tables—featuring a natively managed Iceberg table format with over two million tables available—and S3 Vectors, which added elastic vector indexing for generative AI workloads. S3 Files completes this trifecta by addressing file-oriented applications. AWS is constructing a unified multimodal storage layer, allowing a single object storage service to present interfaces tailored to various consumption modes without forcing teams to choose a specific storage architecture upfront.
This evolution in storage infrastructure is accompanied by simultaneous advancements in security layers. AWS also made two significant cybersecurity announcements aimed directly at security leaders. The first is the rollout of Claude Mythos Preview on Amazon Bedrock, part of Project Glasswing—an initiative highlighted on IT Social on April 8. Developed by Anthropic, this model facilitates autonomous vulnerability detection in highly exposed software, now accessible to security teams via AWS infrastructure. AWS’s Chief Security Officer, Amy Herzog, detailed the operational scale: analyzing 400 trillion network flows daily, blocking 300 million malicious encryption attempts in 2025, and reducing SecOps log analysis time from six hours to just seven minutes through AI—resulting in a fifty-fold gain in processing efficiency.
AWS Security Agent Moves to General Availability
What sets the AWS Security Agent apart from traditional detection tools is its operational design. Rather than merely identifying vulnerabilities, it actively tests them for verification before generating an alert. Each result comes with a CVSS score and detailed reproduction steps. During its preview phase, testing durations decreased from weeks to just hours. For security leaders, this transition to controlled availability alters the fundamental nature of the tool, enabling the generation of an inventory of confirmed, actionable vulnerabilities complete with necessary prioritization and remediation guidelines.
These announcements reflect a broader structural trend within AWS’s cloud offerings. The provider is gradually integrating advanced AI capabilities—whether developed in-house or through partnerships—directly into its infrastructure and security layers, without positioning them as optional modules. For CIOs and security executives, this means that efficiency gains tied to AI become accessible without additional deployment efforts, albeit at the cost of increased dependency on the AWS ecosystem and exposure to the Cloud Act—a consideration that regulated organizations cannot afford to overlook.



