AWS MLHow Standard Bank Group automates AI risk assessment for generative AI
Standard Bank Group uses an AWS pipeline to gather evidence, test generative AI applications, and evaluate them against more than 300 checks derived from external standards and bank policies, cutting reported risk reviews from two weeks to about two hours across six assessments.
CloudflareHow Cloudflare WAF helps protect against phpAfterShell and other critical PHP flaws
PHP’s phpAfterShell flaw exploits interrupted writes to file-based sessions, enabling attacker-controlled data to be deserialized; Cloudflare WAF blocks known attack patterns and adds paid-plan rules for related PHP-FPM and SOAP flaws, but upgrading PHP and restarting workers is the complete fix.
Apple MLDepth Pro 2: Sharp and Temporally Consistent Metric Monocular Depth
Depth Pro 2 uses one scalable network to estimate native-resolution metric depth from images and video, aiming to combine sharp boundaries, temporal consistency, and metric accuracy without the usual trade-off.
OpenAICan AI help scientists identify promising drug candidates?
MedChemBench evaluates how well AI models interpret molecular structures, reason from experimental evidence, and select compounds in early drug discovery.
Publisher summary
OpenAICombatting adversarial distillation
OpenAI’s layered approach to adversarial distillation combines model training, safeguards, detection, and enforcement to protect against capability extraction.
Publisher summary
AWS MLIntroducing enhanced infrastructure observability in Amazon SageMaker HyperPod
Amazon SageMaker HyperPod’s enhanced observability combines 30-second cluster, host, GPU, job, and optional EFA telemetry with peer-based GPU outlier detection and rules for hardware faults, network degradation, throttling, and overheating across EKS and Slurm training clusters.
ElevenLabsElevenAgents for Healthcare: Build an outbound appointment reminder agent with Architect
The article demonstrates building an outbound healthcare reminder agent with ElevenAgents Architect that verifies patients before revealing appointment details and reads or updates appointments directly in an EHR, while testing exposed the need to isolate sensitive data by workflow node and enforce reliable handoffs.
Vercel completes ISO/IEC 42001:2023 certification
Vercel earned ISO/IEC 42001:2023 certification after an independent audit of its AI management system, which covers impact assessments, pre-release evaluations, data controls, and ongoing oversight across its infrastructure and AI products.
Google CloudHow to simplify cloud access with Unified Access Policies, now in preview
Google Cloud’s preview Unified Access Policies combine allow and deny rules in reusable policies that can be bound across multiple organization, folder, or project resources, simplifying centralized least-privilege management and scaling access control for human, workload, and agent identities.
Google CloudHow Vodafone is using the Data Engineering Agent to improve their data pipelines
Vodafone Czech Republic used Gemini-powered Data Engineering Agent to migrate monolithic BigQuery stored procedures into modular Dataform pipelines, enabling parallel execution and cached intermediate tables; the migration cut effort by 90%, doubled execution speed, and reduced costs by 30%.
Google CloudWhat’s new in Managed Service for Apache Spark 4.0
Managed Service for Apache Spark 4.0 brings Spark 4.2 and a shared modern runtime to serverless and managed clusters, with agent-assisted development and troubleshooting, faster Gen4 startups, Lightning Engine acceleration, flexible capacity, and automatic Lakehouse catalog access.
Google CloudDrowning in clicks: Scaling Black Friday pipelines with Managed Airflow agents
Managed Airflow’s console agent diagnoses failures and checks upgrade readiness, while Data Agent Kit helps migrate DAGs; in the example, moving from hourly pandas batches to Airflow 3 event-triggered, per-file processing keeps task memory bounded as traffic grows.