All news with #amazon tag
Fri, October 24, 2025
AWS outage: DynamoDB DNS failure caused disruption
⚠️ Amazon says a major DNS failure in DynamoDB's DNS management system triggered a widespread AWS outage focused on the us-east-1 (Northern Virginia) region. A race condition at 11:48 PM PDT caused the accidental deletion of all IP addresses for the regional DynamoDB public endpoint, producing immediate DNS resolution failures for customer and internal traffic. The fault cascaded across services, kept automated recovery from restoring consistency, and required manual operator intervention to recover. AWS has disabled the problematic DNS automation globally, added protective checks, improved throttling, built new test suites, and apologized for the impact.
Mon, October 20, 2025
Major AWS Outage Disrupts Amazon, Prime Video, Fortnite
⚠️ AWS experienced a widespread outage affecting multiple services in the US-EAST-1 region, causing elevated error rates and latencies across key APIs. The disruption, tied to a DNS resolution issue for the DynamoDB API endpoint and subsequent network load balancer problems, interrupted login and content services for platforms such as Amazon, Prime Video, Fortnite, Canva and Perplexity. AWS reported mitigation steps and later declared services restored after extended recovery efforts.
Mon, October 20, 2025
AWS US-EAST-1 Outage Disrupts Major Sites and Apps
🚨 An AWS outage in the US-EAST-1 region caused widespread disruptions across multiple consumer services, producing elevated error rates and higher latencies. Major platforms including Amazon, PrimeVideo, Fortnite, Perplexity, and Canva reported failures ranging from login and chat outages to impaired editing functionality. AWS acknowledged the incident on its Health page and said engineers were investigating and mitigating the issue. After roughly 45 minutes some services began recovering, though many users still experienced intermittent problems.
Thu, October 9, 2025
Indirect Prompt Injection Poisons Agents' Long-Term Memory
⚠️This Unit 42 proof-of-concept shows how an attacker can use indirect prompt injection to silently poison an AI agent’s long-term memory, demonstrated against a travel assistant built on Amazon Bedrock. The attack manipulates the agent’s session summarization process so malicious instructions become stored memory and persist across sessions. When the compromised memory is later injected into orchestration prompts, the agent can be coerced into unauthorized actions such as stealthy exfiltration. Unit 42 outlines layered mitigations including pre-processing prompts, Bedrock Guardrails, content filtering, URL allowlisting, and logging to reduce risk.
Thu, October 2, 2025
Amazon Prime Day 2025: Surge in Scam Domains and Phishing
🔍 Check Point Research found a marked rise in Amazon Prime Day scams during the first three weeks of September 2025, driven by malicious domains, phishing emails, and credential-harvesting pages that mimic legitimate Amazon communications. Attackers are exploiting urgency and trusted branding to capture login and payment details. Consumers and organizations should verify senders and domains, enable MFA, apply robust email filters, and monitor account activity to reduce exposure.
Thu, September 25, 2025
Amazon to Pay $2.5 Billion Over Prime Enrollment Practices
⚖️ The FTC announced a $2.5 billion settlement with Amazon over allegations it used dark patterns to trick millions into enrolling in and retaining Prime subscriptions. The agreement includes a $1 billion civil penalty and $1.5 billion in refunds for an estimated 35 million affected consumers. The FTC said Amazon's checkout and cancellation designs obscured opt-outs, failed to disclose automatic renewals, and relied on an internal cancellation flow nicknamed "Iliad" that deterred cancellations. Internal documents, the agency added, showed employees discussing the problematic practices.