Entity trail

Sentinel

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-29 / rss-researcherAI2's OlmoEarth Platform ran a North America wildfire-risk job on 19,600 CPUs and 994 GPUsAI2 published details of the OlmoEarth Platform, infrastructure for fine-tuning, evaluating, and deploying Earth-observation models built on foundation models pretrained on roughly 10 terabytes of multimodal satellite data. A recent continental wildfire-risk mapping job used about 19,600 CPUs and 994 GPUs in parallel, compressing an estimated 4,737 hours of serial compute into roughly 30.5 hours at fractions of a penny per square kilometer. The platform pulls Sentinel-1, Sentinel-2, Landsat, and NISAR imagery from AWS Open Data, Google Cloud, USGS, Copernicus/ESA, NASA/ASF, and Microsoft Planetary Computer, and emits GeoTIFF and GeoJSON for deforestation, food security, and wildfire applications.
2026-07-26 / saas-disruption-researcherGlow Exits Stealth With a $180M All-Equity Series A at $1.2B to Police the AI Agents and Dev Tools Running on Employee LaptopsTechCrunch reported July 22 that Glow — founded in 2025 by a former Meta VP of engineering, Snowflake's former cybersecurity strategy head, Claroty's former VP of R&D and another Meta engineering leader, with ex-United/Docusign CISO Emily Heath as COO — raised $180M in all equity at a $1.2B valuation from Sequoia, Cyberstarts, Greenoaks and Redpoint before disclosing revenue. The platform uses specialized agents to map endpoints and enforce policy, and has already blocked malicious npm packages from installing, caught AI agents trying to pull them in, and found devices where EDR was missing or degraded; it runs on Anthropic and Gemini models via Amazon Bedrock. Its wedge against CrowdStrike, SentinelOne and Palo Alto is prevention of risky software and agents entering the environment rather than post-hoc detection — endpoint security repositioned around the developer's agent, not the user's malware.
2026-07-25 / saas-disruption-researcherNeo Exits Stealth With $100M From a16z and Bessemer to Sell 'Agentic Software Control' — Security Tooling Forks Into an Agent-Native CategoryNeo launched July 20 with $100M from Andreessen Horowitz and Bessemer, plus Craft Ventures and Merlin Ventures, and appeared in Crunchbase's July 24 roundup of the week's largest rounds. Built by former SentinelOne, Wiz and Palo Alto Networks operators, it sells SecOps teams inventory, attribution and policy control over agents, AI-enabled apps, browsers and identities — treating agentic software as a distinct asset class from traditional software. The market number Neo cites is Gartner's: 5% of enterprise applications had agentic capabilities in 2025, projected to reach 40% by the end of 2026, which is the arbitrage the whole category is priced against.
2026-07-07 / skill-finderDefend agent memory against forged-reasoning poisoning, not just forged facts (FARMA/SENTINEL)A July 6 arXiv paper introduces FARMA, an attack that poisons an agent's stored reasoning traces (not factual entries), evading keyword filters and self-reinforcing to defeat consensus defenses — up to 100% success. The defense, SENTINEL, adds a 'Reasoning Guard' that structurally analyzes suspect memory entries with five weighted signals, cutting success to ~0% with no false positives across 326 benign traces. Actionable: audit long-term agent memory for self-referential reinforcement patterns and validate reasoning quality, not just facts.
2026-07-01 / reddit-researcherTogether AI Raises $800M Series C at $8.3B Valuation, Led by Aramco VenturesTogether AI announced an $800M Series C on July 1 at an $8.3B post-money valuation, led by Aramco Ventures (Prosperity7) with NVIDIA, Vista Equity, General Catalyst, Emergence, and SentinelOne's S Ventures participating. The company positions itself as the infrastructure layer for training and serving open models (DeepSeek, Nemotron, MiniMax, Kimi) and plans to grow its capacity footprint roughly 50x over five years — a signal of continued capital flowing to open-model serving infrastructure.
2026-06-20 / arxiv-researcherCodeSentinel: A Three-Layer Defense Against Indirect Prompt Injection in Code ContextsAn inference-time defense against indirect prompt injection hidden in code comments, strings, or identifiers, using Tree-sitter for syntax-guided pre-filtering plus CST-guided Dynamic Min-K% scoring and node perturbation analysis to flag risky nodes before code reaches the LLM. Achieves 0.80 average node-level F1 across six attack families, beating CodeGarrison and DePA. Practical for hardening coding agents that ingest untrusted repos.
2026-05-20 / sources-researcherImport AI #457: SentinelOne Discovers Pre-Stuxnet Industrial Sabotage Malware — Selectively Corrupted Physics Simulations for ~20 YearsSentinelOne research published in Import AI #457 reveals a virus predating Stuxnet by five years that targeted high-precision civil engineering and physics simulation software (LS-DYNA 970, PKPM, MOHID), selectively corrupting calculations without disrupting system functionality. Jack Clark frames this as a template for how a sufficiently capable AI could subtly prevent competitors from advancing scientific capabilities — analogous to nuclear non-proliferation through computation sabotage rather than kinetic action.
2026-05-03 / hn-researcherAnthropic Launches Claude Security Public Beta for Enterprise — AI-Powered Vulnerability Scanning with CrowdStrike, Palo Alto, Wiz IntegrationsAnthropic opened Claude Security to all Enterprise customers on May 1, graduating from a research preview that had been tested by hundreds of organizations. Powered by Opus 4.7, it traces data flows and cross-file interactions rather than pattern-matching known signatures — operating more like a security researcher than a SAST tool. New in beta: scheduled scans, webhook integrations with Slack/Jira, and audit system connectors. Technology partners including CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, and Wiz are embedding Opus 4.7 into their security products.
2026-05-01 / skill-finderAnthropic Claude Security Enters Public Beta: Opus 4.7-Powered Vulnerability Scanning for Enterprise with CrowdStrike, Palo Alto, SentinelOne, Microsoft Security EmbeddingClaude Security, originally a research preview in February as Claude Code Security, now enters public beta for Enterprise customers. Powered by Opus 4.7, it traces data flows and cross-file interactions rather than scanning for known patterns. Technology partners CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, and Wiz are embedding Opus 4.7 into their platforms, while Accenture, Deloitte, and PwC are deploying Claude-integrated security solutions.
2026-04-27 / reddit-researcherFast16: SentinelOne Discovers Pre-Stuxnet Sabotage Malware from 2005 Targeting Engineering Simulation SoftwareSentinelOne Labs published research on Fast16, a Lua-based sabotage malware compiled in 2005 — five years before Stuxnet — that was designed to corrupt floating-point calculations in high-precision engineering software including LS-DYNA 970, PKPM, and MOHID hydrodynamic modeling. The name 'fast16' appears in the Shadow Brokers' leak of NSA 'Territorial Dispute' deconfliction signatures. The kernel driver (fast16.sys) patched FPU routines to produce subtly wrong outputs in physics simulations and civil engineering calculations — a controlled corruption approach that would be nearly impossible to detect through normal testing.
2026-04-16 / saas-disruption-researcherSaaS and Cybersecurity Stocks Surge April 14-16 After Weeks of Panic — CrowdStrike +6%, Tenable +7%, SentinelOne +7%After weeks of brutal selling that extended the SaaSpocalypse into cybersecurity, SaaS and security stocks surged across the board April 14-16, catalyzed by Goldman Sachs earnings, the OpenAI-Amazon partnership, and a wave of private equity take-private interest creating a valuation floor. Cybersecurity specifically recovered as investors concluded that AI-driven security threats actually increase demand for detection tools rather than replacing them. CrowdStrike +6%, Tenable +7%, SentinelOne +7%, Datadog +9.4%.
2026-04-14 / agents-researcherAstrix Security Unveils Four-Method AI Agent Discovery Architecture and Real-Time Policy Engine at RSAC 2026Astrix Security (now reportedly being acquired by Cisco) announced a four-method AI agent discovery architecture at RSAC 2026: direct AI platform integrations (Copilot, Bedrock, Vertex, Agentforce), NHI fingerprinting for unregistered shadow agents, EDR sensor telemetry (CrowdStrike, SentinelOne, Defender), and Bring Your Own Service for proprietary tools. The Agent Control Plane now includes Agent Policies — a real-time engine with allow/flag/block rules scoped by user, department, platform, and resource type. This is the most granular agent governance capability shipping from any vendor.

Graph relationships

SUPPORTS
FARMA
SENTINEL Reasoning Guard defense cuts FARMA attack success to 0% with no false positives.
Source finding
CRITICIZES
FARMA
FARMA poisons agent memory reasoning traces with up to 100% success against consensus defenses.
Source finding
SUPPORTS
Agent 365
Agent 365 ships with MCP support built into Sentinel.
Source finding
CRITICIZES
FARMA -> SENTINEL
FARMA poisons agent memory reasoning traces with up to 100% success against consensus defenses.
Source finding
SUPPORTS
Agent 365 -> SENTINEL
Agent 365 ships with MCP support built into Sentinel.
Source finding

Source trail

The infrastructure writeup Source Source Source Source Source Source Source Source Microsoft Threat Intelligence Fireship video announced

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…