Version history

AWS Bedrock updated its Service Terms to add AWS Elemental Inference to the list of Indemnified Generative AI Services and removed specific restrictions on the Amazon Mechanical Turk workforce option for Amazon Augmented AI. The terms also reorganized section numbering in the SageMaker AI section. These changes expand the scope of services covered under AWS's generative AI indemnification and simplify workforce data handling requirements.

AWS Bedrock updated its Service Terms document date from September 1, 2026 to September 11, 2026. The only substantive change detected is the removal of the word 'Preview' from the Amazon Bedrock AgentCore Payments feature description, indicating the feature has moved from preview status to general availability. This means the AgentCore Payments service is now a standard offering rather than an experimental or limited-access feature.

AWS Bedrock updated its AWS Service Terms to expand the list of services covered under its Indemnified Generative AI Services definition. The updated terms add three new services to the list: AWS DevOps Agent, AWS Security Agent, and Amazon Quick. This change means these three services now fall under AWS's indemnification framework for generative AI output, which may affect how liability and intellectual property claims are handled for users of these services.

AWS Bedrock updated its Service Terms to introduce a new 'AWS (new)' sign-up option available from August 17, 2026, which organizes accounts into AWS projects within an AWS organization structure. The updated terms establish new rules for project owners and team members, including automatic disclosure of billing and profile information, shared content licensing, and spend limit controls that can suspend or permanently close accounts. The terms also removed language about free access to Amazon Bio Discovery without a payment method and added comprehensive spend limit provisions that allow AWS to suspend or close accounts if spending thresholds are exceeded.

AWS Bedrock updated its Service Terms to clarify how it handles customer regulated data in professional services engagements. Previously, the terms stated that customers should not disclose regulated data to AWS personnel except as directed and stated in a Professional Services Order. The updated language adds a new provision: if a customer expressly directs AWS to access regulated data in a Professional Services Order and applicable data protection laws apply, the AWS Professional Services Supplementary Data Processing Addendum (DPA) is automatically incorporated into the agreement. This establishes a formal data processing framework for regulated data handling in professional services contexts.

AWS updated its Service Terms to clarify safety and liability provisions for AWS IoT services and to expand the definition of AI Services to include AWS IoT SiteWise Scenario Discovery. The updated terms add explicit restrictions stating that AWS IoT SiteWise Scenario Discovery is not designed for real-time vehicle control or regulatory submissions, and users may not rely on its data to determine vehicle safety or compliance. The terms also removed AWS IoT 1-Click and AWS IoT Events from the AWS IoT Services definition while adding AWS IoT SiteWise Scenario Discovery.

AWS updated its EC2 pricing and reservation terms, effective July 17, 2026. The company introduced new restrictions on On-Demand Capacity Reservations, prohibiting resale to third parties and reserving the right to cancel purchases or terminate instances if resale is suspected. AWS also revised termination windows for Capacity Blocks for ML, extending the notice-free termination period from 30 minutes to 60 minutes for UltraServer instance types. Additionally, AWS rebranded its Amazon Sidewalk qualification program from 'Works with Amazon Sidewalk' to 'On Amazon Sidewalk' and simplified trademark licensing requirements for qualified devices.

AWS Bedrock's Service Terms replaced its AWS IQ marketplace section with new provisions governing AWS Config's use of configuration data for service improvement. The previous terms required exclusive use of AWS IQ for 24 months when connecting with a Provider and released AWS from liability for Provider Services. The updated terms authorize AWS to use and store Config Content to develop and improve AWS Config and related security services, with an option to store data outside the customer's configured region, but permit customers to opt out via an AWS Organizations AI services policy.

AWS added a new section (110) to its Service Terms establishing terms for free exploratory technical consulting and advisory services. The updated terms specify that AWS may provide free services including technical assistance, prototypes, and AI engagements, but may discontinue them without notice. The terms establish that users are solely responsible for testing, deployment, and compliance of any deliverables, including AI solutions, and that AWS retains rights to develop similar products based on content created during these engagements.

AWS updated its Service Terms to add Amazon GuardDuty to the list of services that may incorporate generative AI features powered by Amazon Bedrock. The updated terms now state that Amazon GuardDuty, along with 26 other named services, may use prompts to generate output, and that Amazon Bedrock abuse detection mechanisms may apply to these services. This expands the scope of services subject to Bedrock's AI capabilities and associated abuse detection.

AWS Bedrock updated its abuse detection and content retention practices. Previously, the terms stated that automated abuse detection mechanisms "may apply" to services using Bedrock. The updated language specifies that Bedrock may store service inputs and outputs for up to 30 days to detect policy violations, and clarifies that if child sexual abuse material is detected, AWS may review the flagged content before reporting to authorities. The change also adds a reference to "applicable Service-specific documentation" for additional details and removes a reference to "related information sharing with third parties."

AWS Bedrock restructured its Amazon RDS service terms, reorganizing sections for RDS, RDS Custom, and database engine-specific provisions. The updated terms add new sections on Trusted Language Extensions, Engine End of Life (EOL) management, and Oracle AI Database@AWS, while clarifying licensing requirements for RDS Custom with Microsoft, IBM Db2, and Oracle software. Key additions include AWS's right to scan extension code for security purposes, responsibility for maintaining EOL engine versions, and requirements for License Mobility verification when using Microsoft software via Bring Your Own Media (BYOM).

AWS Bedrock's Service Terms were updated to add comprehensive governance requirements for Amazon Connect Talent, a new AI service for employment-related decisions. The updated terms establish that customers must provide privacy notices to job applicants, obtain necessary consents, monitor outputs for bias and discrimination, maintain audit records, and ensure compliance with labor, employment, anti-discrimination, and data privacy laws. The terms also clarify that AWS's bias mitigation efforts do not relieve customers of independent responsibility for preventing discrimination in their use of the service.

AWS Bedrock updated its Service Terms to clarify that Reserved Cache Nodes and Amazon DynamoDB Reserved Capacity purchases are noncancellable and remain owed for the full selected term even if the service agreement is terminated. The terms also expanded data retention practices for Kiro Free Tier users, permitting AWS to store user inputs for up to 60 days for abuse detection and service improvement purposes.

AWS Bedrock added a new section governing Claude Platform on AWS, a service that integrates Anthropic's Claude AI model into AWS infrastructure. Users of Claude Platform on AWS must now comply with Anthropic's separate terms of service, data processing addendum, and usage policy. The updated terms explicitly state that content and metadata provided to Claude Platform on AWS will be processed by Anthropic outside of AWS, and AWS is authorized to transfer this data to Anthropic for processing.

AWS Bedrock added a new service called AgentCore Payments (Preview) that enables developers to build payment solutions using AI agents. The service transmits API calls between AI agents, third-party wallet providers, and sellers, but AWS explicitly states it does not provide regulated financial services, does not hold funds, and is not responsible for wallets, digital assets, or wallet provider services. Developers using AgentCore Payments must comply with applicable laws, obtain necessary licenses, implement safeguards like human-in-the-loop controls, and bear sole responsibility for transactions, disputes, and payment liability.