COGNOSCERE Daily Tech Review — Issue T195 · Saturday, August 1, 2026

Or visit Intelligence Overview for deeper analysis.

PREPAREMicrosoft Copilot AI Worm Is Propagating Through Word Documents via Hidden Prompt Injection

Security researcher Håkon Måløy disclosed a cross-domain prompt injection vulnerability in Microsoft Copilot for Word in which hidden instructions embedded in documents can survive into AI-generated output and self-propagate to newly created files across enterprise workflows such as SharePoint, Teams, and email. Microsoft deployed partial mitigations over a 144-day coordinated disclosure but acknowledges no robust fix exists for the underlying vulnerability class, as current LLMs cannot reliably distinguish untrusted data from executable instructions.

CSO Online · Cybersecurity / Enterprise Software · Relevance: 0.9 · Source →

prompt injection, AI worm, Copilot, Microsoft 365, enterprise security, document integrity, self-propagating malware

PREPAREOkta Acquires AI Security Startup Permiso for ~$200M to Guard Against AI Agent Threats

Okta agreed to acquire Permiso Security for approximately $200 million in an almost all-cash deal, gaining technology that monitors behavior of employees, service accounts, applications, and AI agents after authentication to detect identity threats in real time. The acquisition expands Okta's security posture from login-time verification to continuous post-access behavioral monitoring across cloud and SaaS environments.

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

identity security, AI agents, machine identity, acquisition, cloud security, ITDR, non-human identities

WATCHMajority of Corporate IT Workloads Now Off-Premises for the First Time

Uptime Institute's survey of more than 800 operators found that third-party facilities now host 46% of enterprise workloads, surpassing company-owned datacenters at 44% — the first time off-premises infrastructure has taken the lead in the study's history. Typical rack density also climbed above 11 kW as organizations replace older servers with more power-intensive hardware to support AI workloads.

The Register · Cloud Computing / Data Infrastructure · Relevance: 0.8 · Source →

cloud computing, data center, enterprise IT, off-premises, infrastructure, colocation, rack density

WATCHCisco Releases AI Models for Deep Networking Operations with Cloud Control Agentic Platform

Cisco is preparing to release AI models specifically trained on routing, switching, and network troubleshooting tasks, accompanied by its Cloud Control agentic operations platform that lets administrators diagnose issues across Cisco products through a single interface. The platform explains the potential impact of administrative actions such as reboots, supports on-premises deployment, and is planned for US availability by end of August 2026.

The Register · Networking / Enterprise AI · Relevance: 0.8 · Source →

Cisco, networking AI, agentic operations, network management, enterprise IT, routing, switching

ACTEU AI Act Enforcement and Mandatory Transparency Rules Activate August 2, 2026

Starting August 2, 2026, the EU AI Office and national authorities gained active enforcement powers over the AI Act, including the ability to request model documentation, conduct evaluations, and impose fines of up to €15 million or 3% of global turnover for non-compliance; Article 50 transparency obligations requiring chatbot disclosure, deepfake labeling, and machine-readable AI content marking also become enforceable on this date. High-risk AI system obligations were postponed to December 2027 and August 2028 via the AI Omnibus, but GPAI model enforcement activates immediately.

European Commission · AI Regulation / Governance · Relevance: 0.9 · Source →

EU AI Act, AI regulation, transparency, deepfakes, GPAI, compliance, AI governance, enforcement

PREPAREOpenAI Cuts GPT-5.6 Luna Price by 80% and Terra by 20% Amid Competitive AI Price War

OpenAI slashed the price of its GPT-5.6 Luna model by 80% to $0.20 per million input tokens and Terra by 20% to $2 per million input tokens just three weeks after the family launched, citing efficiency gains partially achieved by its Sol model rewriting its own GPU inference code. The aggressive cuts respond to enterprise cost sensitivity and competition from Chinese open-weight models including DeepSeek and Kimi, which have captured substantial US enterprise token market share.

CNBC · AI Models / Enterprise AI Adoption · Relevance: 0.9 · Source →

AI pricing, frontier models, enterprise AI, AI cost, token economics, AI competition, Chinese AI models

WATCHOpenAI Publishes EU Compliance Statement Ahead of AI Act Enforcement Activation

OpenAI published a detailed compliance narrative endorsing the EU's General-Purpose AI Code of Practice and the Code on Transparency of AI-Generated Content, describing its Preparedness and Frontier Governance Frameworks and its EU Cyber Action Plan with European cyber agencies ahead of August 2 enforcement activation. Observers noted the statement omits the Code's Copyright chapter, which requires a public training data summary — a gap that becomes subject to EU enforcement scrutiny starting August 2.

OpenAI · AI Regulation / Governance · Relevance: 0.8 · Source →

EU AI Act, AI regulation, AI governance, compliance, GPAI Code, copyright, AI transparency, OpenAI Europe

PREPARESnowflake Launches Cortex AI Gateway to Track Agent Actions, Enforce Policies, and Control AI Costs

Snowflake unveiled Cortex AI Gateway, a centralized runtime control plane built on technology from its Natoma acquisition that governs how AI agents access enterprise models, tools, MCP servers, and data while providing finance and IT teams with per-agent token cost attribution and spending limit enforcement. The launch includes identity security integrations with 1Password, SailPoint, Aembit, and Okta to enable task-scoped agent permissions and end-to-end audit trails across first- and third-party agents.

CIO.com · AI Governance / Analytics Infrastructure · Relevance: 0.9 · Source →

AI governance, agentic AI, FinOps, MCP, enterprise AI, cost control, agent observability, data governance

WATCHMicrosoft and Databricks Extend Strategic AI Partnership Through the 2030s

Microsoft and Databricks announced an expansion of their partnership extending into the 2030s, with Databricks committing to run its own core operations on Azure Databricks and deepening integration across Microsoft 365, Teams, Copilot, Power BI, and Microsoft Purview to bring governed, real-time data and AI into business workflows. Databricks also moves to Azure Cobalt 200 Arm-based infrastructure, promising up to 50% performance improvement for agentic workloads.

Microsoft · Analytics Infrastructure / Enterprise AI · Relevance: 0.8 · Source →

data lakehouse, enterprise AI, analytics, Microsoft Azure, Databricks, agentic AI, data governance, partnership