TLDR IT 2026-05-27

Israeli researchers attributed a March cyberattack on LA Metro to an Iran-linked hacking group, saying about 700GB of internal data was stolen. Transit service reportedly kept running, but customer-facing systems like arrival information and fare-card functionality were disrupted.

Dropbox founder Drew Houston is stepping down as CEO, with core-business GM Ashraf Alkarmi becoming co-CEO and eventual successor. The transition comes as Dropbox tries to reposition around AI products like Dash and bring in new product leadership.

Enterprise storage vendors are adjusting their products and architectures in response to a broader memory crunch. Constrained RAM supply and pricing impact storage system design and performance, and vendors are adopting different strategies to deal with the situation. Memory availability has become a key constraint for enterprise storage deployments.

Most companies build AI SDRs incorrectly by treating them as tools for better cold emails rather than full systems that run the entire outreach-to-qualification workflow. Current offerings often automate messaging without verifying lead qualification, maintaining continuity, or taking effective action with accurate internal information. Krazimo approaches AI SDRs as a connected, monitored, end-to-end system with onboarding, context, and tool access to reduce delays and performance drift.

Modern AI-driven systems are getting harder to manage, with new layers introducing blind spots, rising costs, and unpredictable behavior. This Dynatrace guide breaks down how teams gain visibility into models, prompts, and pipelines to improve reliability and control spend. Read the guide

Cohere acquired Reliant AI to deepen its sovereign enterprise AI push in healthcare, life sciences, and other regulated industries. The deal brings in Reliant's biomedical datasets, domain-focused technology, and research team as Cohere tries to verticalize beyond general enterprise AI.

Anthropic introduced 28 integrations for Claude across security, compliance, identity, DLP, SIEM, and AI governance platforms, letting enterprise teams monitor Claude activity through existing security workflows and dashboards. The move signals that AI assistants are increasingly being treated like governed enterprise SaaS platforms rather than standalone productivity tools.

Google Agent Executor is an open source runtime designed to help enterprises run AI agents reliably at scale in production. It adds durable, resumable execution for long-running workflows, secure sandboxing, connection recovery, and controls for distributed session consistency, with features like trajectory branching for testing alternate execution paths.

OpenAI hired ServiceNow CMO Colin Fleming to lead business marketing, replacing Kate Rouch. It's a personnel move, but relevant because it signals OpenAI's continued enterprise go-to-market push.

Salesforce describes an enterprise agent governance approach that shifts protection from individual agents to a unified platform layer across identity, data, and APIs. It explains how identity is propagated through multi-system workflows, how data access is centrally enforced regardless of agent behavior or request path, and how API and AI model governance are integrated into one control model to maintain security, auditability, and trust.

Tori is a small, secure terminal UI (TUI) agent that monitors Docker and hosts by collecting metrics and tailing logs, running continuously on a server.

Google is expanding external testing for CodeMender, its AI agent for finding and fixing software vulnerabilities.