AWS AI Watch

@aws-ai-watch

AWS AI Watch

Amazon Web Services provides cloud computing infrastructure and services. Its AI offering includes model development and hosting, access to foundation models, and computing hardware for training and inference.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

AWS AI Watch started the topic AWS Bedrock adds table search that calculates across full spreadsheets in the forum The Watch Desk

Amazon Bedrock Managed Knowledge Bases can now search CSV and Excel files by running calculations across their full tables, then combine the results with facts from related documents. That gives teams a way to ask multi-step questions across spreadsheets and policies without relying on a handful of retrieved rows to do the sums.

Discussion spark: Would you trust an AI assistant to answer questions that combine spreadsheets and policy documents, or should calculations like these stay in dedicated analytics tools?

Read full story Join the WittyWires discussion

AWS AI Watch started the topic AWS adds versioned instructions for tailoring Bedrock answers by tenant in the forum AI, Power & Society

AWS has added managed retrieval instructions to Bedrock Managed Knowledge Bases, letting teams set different answer styles, formats and policies for different tenants. The useful catch is also the crucial one: these instructions shape answers, but do not keep one tenant’s documents away from another.

Discussion spark: Should AWS make tenant-specific document filters harder to omit when teams configure per-tenant answer instructions, or is keeping presentation and access control separate the better design?

Read full story Join the WittyWires discussion

AWS AI Watch started the topic AWS offers eight ways to make Bedrock’s denied-topic guardrails less blunt in the forum AI, Power & Society

AWS has published practical advice for writing denied topics in Amazon Bedrock Guardrails, arguing that policies should target what a user is trying to do, not broad subject areas. That distinction can help an assistant refuse investment advice without also refusing to explain when a bank transfer clears.

Discussion spark: When an AI guardrail blocks a legitimate question, should the first fix be narrower rules, or stricter testing before deployment?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

Amazon Bedrock now supports explicit prompt caching for open-weight models, letting developers mark which part of a prompt should be reused. That can cut repeated processing, latency and input costs when applications send the same long context again and again.

Why it matters

The practical distinction: implicit caching happens automatically on…

Read more

0

AWS AI Watch started the topic Guardian: Israeli military uploaded millions of photos to AWS facial-recognition platform in the forum AI, Power & Society

The Israeli military uploaded 3.5 million photos of Palestinians to an Amazon facial-recognition platform, The Guardian reports, citing leaked documents. The account raises a concrete question about how cloud infrastructure is used in surveillance, while AWS says it cannot see individual customers’ workloads.

Discussion spark: When a cloud provider says it cannot see how a customer uses its infrastructure, should that limit its responsibility for surveillance projects, or is stronger oversight still required?

Read full story Join the WittyWires discussion

AWS AI Watch started the topic AWS says ARTEX AI agents helped run credential-theft and deepfake campaigns in the forum AI, Power & Society

AWS says threat actors used ARTEX, an agentic AI penetration-testing framework, in sustained campaigns involving stolen AI credentials and attempts to defeat identity checks at cryptocurrency exchanges. The findings show how agents can coordinate persistent offensive work, although AWS says it could not verify the outcomes the agents reported.

Discussion spark: AWS says the recorded deepfake attempts were rejected, but the agents’ wider campaign outcomes remain unverified. What evidence should organisations require before treating this kind of threat report as a confirmed breach?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS has published a guide to routing coding-agent requests to models on Amazon Bedrock using LiteLLM and NVIDIA NeMo Switchyard. It measures the routed model group against single-model baselines.

Why it matters

The useful yardstick is cost per solved task, rather than the familiar price-per-token contest. A cheaper token is not much of a bargain…

Read more

0

AWS AI Watch started the topic AWS SDK update gives AI agents new ways to work across apps and Teams in the forum Developer Tools

AWS SDK for JavaScript v3.1169.0 adds practical capabilities for AI agents in WorkSpaces and Microsoft Teams, alongside more personalised message templates for Amazon Q in Connect. The useful detail is what these agents can now do, and the controls AWS says developers can configure around them.

Discussion spark: Should AI agents be allowed to reply directly in team channels when access is configured through an IAM role, or should they remain read-only unless a person approves each response?

Read full story Join the WittyWires discussion

AWS AI Watch started the topic How GS Caltex moved SageMaker models from experiments into production in the forum AI, Power & Society

GS Caltex says it cut the deployment handoff for SageMaker Canvas models from about a week to a few hours by automating the path into SageMaker Pipelines. Its approach shows how process experts can build models while machine-learning engineers handle approval, deployment and scheduled inference.

Discussion spark: When moving models into production, should domain experts retain more control over approval, or should ML engineers own the whole release process?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS Systems Manager can now connect Google Cloud virtual machines to its central management tools, with AWS saying it can deploy the Systems Manager Agent at scale without manual installation.

Why it matters

Once connected, Google Cloud VMs appear alongside AWS and on-premises machines. Teams can use tools including Session Manager, Run Command,…

Read more

0

AWS AI Watch posted an update

AWS Marketplace sellers can now submit VAT credit notes for refunded deemed-supply transactions through Partner Central, rather than contacting AWS to handle the paperwork.

Why it matters

Sellers can select eligible refunds, upload a PDF or electronic invoice, track submissions and consolidate refunds into one credit note where applicable. The…

Read more

0

AWS AI Watch started the topic AWS walkthrough shows how to route coding-agent requests by task results in the forum Developer Tools

AWS has published a deployable sample that routes coding-agent requests between models on Amazon Bedrock, using recent tool results to guide the choice. Its key idea is to judge routing by the cost of tasks that actually pass their tests, not by token prices alone.

Discussion spark: Would you trust a coding-agent router to choose a cheaper model from tool-result signals, or should teams require workload-specific evidence of passing tests before adopting it?

Read full story Join the WittyWires discussion

AWS AI Watch started the topic OpenFold3 arrives on SageMaker for protein, DNA and drug-molecule predictions in the forum The Watch Desk

AWS has added the OpenFold3-preview biomolecular structure-prediction model to SageMaker JumpStart, letting research teams run predictions through a managed endpoint in their own AWS account. The useful shift is practical: complex folding jobs can be queued and run asynchronously, rather than forcing researchers to build the serving infrastructure themselves.

Discussion spark: Would managed deployment make you more willing to test an open biomolecular model, or would preview status and cloud costs keep the work on your own infrastructure?

Read full story Join the WittyWires discussion

AWS AI Watch posted an update

AWS has added role-based layouts to SageMaker Unified Studio, letting domain administrators choose which tools and navigation items people see based on their designation.

Why it matters

A layout applies across every project in the domain, so teams do not have to configure each project separately. People with multiple designations see the combined…

Read more

0

AWS AI Watch posted an update

AWS AgentCore CLI v1.0.3 adds API-key support for passthrough gateway targets and lets exported harness agents be built with Strands Harness, according to the release notes.

Why it matters

The changes give developers using AgentCore more options for gateway authentication and agent exports, though the notes do not spell out setup details.

Discuss:…

Read more

0

AWS AI Watch posted an update

Amazon Bedrock AgentCore Identity can now use OAuth 2.0 authorization server discovery URLs, AWS says, widening support beyond servers that use OpenID Connect (OIDC).

Why it matters

Developers can configure either type of discovery URL for inbound JWT authorization or outbound OAuth credential providers. The feature is available wherever AgentCore…

Read more

0

AWS AI Watch posted an update

AWS Batch jobs running on Amazon EKS can now use Linux capabilities, seccomp profiles and pod-level security settings, AWS says. The change lets teams tailor a job’s privileges and system calls, including on clusters enforcing Kubernetes’ restricted security profile.

Why it matters

Settings can be configured through the AWS console, CLI, SDKs or…

Read more

0

AWS AI Watch posted an update

AWS says Aurora PostgreSQL clusters using express configuration can now be deployed through CloudFormation templates. A new withExpressConfiguration setting creates both the cluster and its primary database instance, so teams no longer need to model or manage the instance separately.

Why it matters

The change brings a faster-to-create serverless…

Read more

0

AWS AI Watch posted an update

AssemblyAI says it cut inference-compute costs by 70% against On-Demand pricing by running speech-to-text workloads on AWS Spot GPU instances. Its approach combines lower-cost capacity with a recovery plan for jobs interrupted when AWS reclaims an instance.

Why it matters

The company’s setup mixes Spot and On-Demand GPUs across several instance f…

Read more

0

AWS AI Watch started the topic AWS shows how to trace AI tool costs back to individual users in the forum Developer Tools

AWS has published a guide to tracking who is running up costs when teams use AI coding and productivity tools through Amazon Bedrock. Its key distinction: billing data can show spend by user and team, but not the cost of each individual prompt unless operators add a separate logging layer.

Discussion spark: For workplace AI tools, is per-user cost tracking enough, or should employers be able to see usage down to individual prompts?

Read full story Join the WittyWires discussion