COGNOSCERE Daily Tech Review — Issue T181 · Saturday, July 18, 2026

Or visit Intelligence Overview for deeper analysis.

ACTCISA Urges Immediate SharePoint Hardening as Exploits Mount

CISA confirmed active exploitation of four on-premises SharePoint Server vulnerabilities — including a CVSS 9.8 zero-day (CVE-2026-58644) added to the Known Exploited Vulnerabilities catalog on July 16 — enabling remote code execution, IIS machine-key theft, and persistent malware deployment across all supported SharePoint versions. Federal civilian agencies were given until July 19 to patch, and CISA directed all organizations to apply July Patch Tuesday updates, enable AMSI in full mode, rotate IIS machine keys after hunting for artifacts, and block direct internet exposure of SharePoint servers.

CISA · Cybersecurity · Relevance: 1.0 · Source →

sharepoint, zero-day, remote-code-execution, patch-tuesday, KEV, ransomware, enterprise-security, IIS

ACTMicrosoft Patches Record 622 CVEs: Active SharePoint and AD FS Zero-Days Demand First Action

Microsoft's July 2026 Patch Tuesday, the largest in company history at 622 CVEs, included two actively exploited SharePoint zero-days and an AD FS authentication bypass (CVE-2026-56155) that, when chained with a remote code execution flaw, creates a path for attackers to forge tokens across an entire federated estate. SharePoint 2016 and 2019 also reached end of extended support on the exact day the patches shipped, compounding risk for organizations still running those versions.

TechTimes · Cybersecurity · Relevance: 0.9 · Source →

patch-tuesday, zero-day, AD-FS, identity, sharepoint, end-of-support, CVE, enterprise-patching

ACTGrok's CLI Caught Uploading All Your Local Files to the Cloud

Researchers found that the Grok Build CLI (from xAI) was uploading entire Git repositories — including full commit histories and sensitive files — to xAI-managed cloud storage, even when users had disabled model-improvement settings; xAI subsequently stopped the uploads via a server-side switch. The incident raises enterprise governance concerns about whether stated data controls in AI coding agents can be trusted, with implications for IP protection and compliance in software development environments.

TLDR IT · Cybersecurity / AI Security · Relevance: 0.8 · Source →

AI-security, data-exfiltration, coding-agent, enterprise-governance, IP-protection, xAI, Grok, privacy

PREPAREBacked by $60M in Funding, Oak Steps Out of Stealth to Fix the Identity Mess That AI Agents Are Making Worse

Israeli identity startup Oak emerged from stealth with a $60M seed round co-led by Accel, CRV, and Greylock Partners to build an AI-native Identity Operating System that governs human, machine, and AI agent identities from a single control plane, removing unnecessary permissions in real time rather than in periodic reviews. The company, founded by serial entrepreneur Shai Morag (who previously sold Ermetic to Tenable for $265M), already has its product generally available and deployed in enterprise customers.

TechCrunch · Cybersecurity / Identity & Access Management · Relevance: 0.9 · Source →

identity, IAM, AI-agents, non-human-identities, seed-funding, zero-trust, cybersecurity, stealth-exit

PREPAREThe 2026 State of SaaS Report

BetterCloud's survey of 525 IT and security professionals found SaaS stacks are growing again, driven largely by AI, with mid-market companies seeing their average app count jump 41% in one year; only 56% of apps in use are IT-approved, while 62% of IT leaders report that manual work prevents strategic projects and 90% still lack true cross-app automation. The findings underscore a widening governance gap as shadow AI and unsanctioned SaaS proliferate.

BetterCloud · SaaS / IT Governance · Relevance: 0.8 · Source →

SaaS, shadow-IT, governance, AI-adoption, IT-security, app-sprawl, automation, mid-market

PREPAREOpen-Source Kimi K3 Closes the Frontier Gap

Chinese AI startup Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model with a 1-million-token context window that blind Arena evaluators ranked first for front-end coding — ahead of Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol — and placed fourth overall on the Artificial Analysis Intelligence Index. Open weights are scheduled for release on July 27, and the model is already available via API at pricing comparable to mid-tier closed models, intensifying competitive and geopolitical pressure on U.S. AI labs.

Axios · AI Models / Open Source · Relevance: 1.0 · Source →

open-source, open-weight, frontier-AI, China, Kimi, MoE, AI-competition, geopolitics, enterprise-AI

PREPAREAnthropic, Blackstone Bet the Next Trillion-Dollar AI Business Is Implementation, Not Models

Anthropic and Blackstone-backed venture Ode with Anthropic — built from the acquired startup Fractional AI and staffed by 100 engineers — is positioning itself as a $1.5B AI implementation firm focused on deploying AI engineering inside enterprises rather than solely improving models, competing directly with OpenAI's deployment business and consulting firms' forward-deployed teams. The venture will follow a 'Claude-first' approach but will also use rival tools when needed, signaling that the integration and deployment layer is emerging as a distinct and high-value segment of the AI market.

TechCrunch · Enterprise AI Adoption · Relevance: 0.9 · Source →

enterprise-AI, AI-implementation, consulting, Claude, deployment, Blackstone, Anthropic, forward-deployed-AI

PREPAREA Scorecard for the AI Age

OpenAI CFO Sarah Friar published a practical AI ROI scorecard that reframes how enterprises should measure AI investment returns — shifting from cost-per-seat to metrics including useful work completed, cost per successful task, dependability, and return on compute. The framework is aimed at business leaders navigating AI budget decisions as usage-based agentic pricing replaces predictable per-seat models.

OpenAI · Enterprise AI Adoption / AI Economics · Relevance: 0.8 · Source →

AI-ROI, AI-metrics, enterprise-AI, agentic-AI, cost-management, AI-adoption, scorecard, CIO

PREPAREAtlassian Extends AI Reach of Jira into Agentic Engineering Workflows

Atlassian is turning Jira into a hub for managing agentic software development by integrating with Claude Code, Cursor, GitHub Copilot, and soon OpenAI Codex, with new capabilities that convert tickets into review-ready pull requests, generate technical specifications, automate background development tasks, and track AI tool and token spending across projects. The update signals a broader shift in how enterprise project management tools are positioning themselves as orchestration layers for AI coding agents.

DevOps.com · Agentic AI / Enterprise AI Adoption · Relevance: 0.8 · Source →

agentic-AI, Jira, DevOps, AI-coding, enterprise-software, AI-spend-tracking, software-development, MCP

WATCHThinking Machines Releases First Open Model — Inkling

Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released Inkling, its first model — a 975-billion-parameter open-weights multimodal system available on Hugging Face under Apache 2.0, with enterprise customization available through its Tinker fine-tuning service. Inkling is designed to balance cost against performance rather than compete for benchmark crowns, and uses roughly one-third the tokens of Nvidia's Nemotron 3 Ultra for equivalent coding performance.

Thinking Machines Lab · AI Models / Open Source · Relevance: 0.9 · Source →

open-weights, Inkling, multimodal, fine-tuning, enterprise-AI, US-open-source, Mira-Murati, agentic-AI