Microsoft is strengthening security around AI agents in Microsoft 365 E7 with new controls including Dynamic Permissions, Dynamic Agent Risk, and Dynamic Conditional Access. The update also expands agent discovery to 60 types, covering tools like Claude Code, OpenAI Codex, and GitHub Copilot, giving security teams better visibility into how agents access data and take actions.
Key Takeaways
- <ul><li>Microsoft adds Dynamic Permissions, Dynamic Agent Risk, and Dynamic Conditional Access for AI agents in Microsoft 365 E7</li><li>Agent discovery now covers 60 types including Claude Code, OpenAI Codex, and GitHub Copilot</li><li>Alphabet posted a $9.07 billion free cash flow deficit in Q3 2026 with $51.4 billion in capital expenditures</li><li>Google Cloud revenue grew 103% and Alphabet has a $561 billion backlog according to Goldman Sachs</li><li>UK government announces £600 million AI investment plan including £200 million in OpenAI and Anthropic model credits</li><li>AM Intelligence plans to raise $3 billion in debt to purchase Nvidia's next-generation Rubin GPUs</li><li>SentinelOne launches Singularity Agentic Detection and Response supporting agents from Salesforce, Microsoft, OpenAI, and Anthropic</li><li>SafeBreach reports one in four malicious breaches is now AI-enabled per IBM's 2026 Cost of a Data Breach Report</li><li>New Reinforced Adversarial Learning method trains reasoning LLMs using a co-trained judge to turn reference answers into rewards</li><li>Julius Baer CIO says AI remains a strong investment but warns single-stock bets carry significant risk</li></ul>
Microsoft adds new security controls for AI agents in 365
Microsoft introduced new security features for AI agents in Microsoft 365 E7. The update includes Dynamic Permissions, Dynamic Agent Risk, and Dynamic Conditional Access. It also expands agent discovery to 60 types, including Claude Code, OpenAI Codex, and GitHub Copilot. The goal is to give security teams more oversight of how AI agents access data and take actions.
Akeneo expands AI agents for product data with enterprise controls
Akeneo launched new AI-powered product data tools and enterprise security controls. The update adds the Ziggy orchestrator integration with Shopify and AI-assisted Product Model Suggestions. It also introduces governance features like Just-in-Time provisioning, SCIM, and SIEM connectivity. Akeneo also added transparent AI labelling to show which content was generated by AI.
SentinelOne launches agentic detection and response for AI security
SentinelOne launched Singularity Agentic Detection and Response, a new security offering for AI agents. It was announced on October 29, 2026, at OneCon 2026 and is available in early access. The tool helps enterprises discover AI agents, monitor their activity, and block threats like prompt injection. It supports agents from Salesforce Agentforce, Microsoft Copilot Studio, OpenAI ChatGPT, and Anthropic Claude.
Alphabet reports a $9 billion free cash flow deficit amid AI spending
Alphabet posted negative free cash flow of $9.065 billion in Q3 2026. Capital expenditures reached $51.355 billion, while operating cash flow fell to $42.29 billion. Revenue rose 24.65% to $127.577 billion, but GAAP EPS missed expectations at $2.12. The company also reported $924 billion in future purchase commitments.
Goldman Sachs still backs Alphabet despite earnings miss
Goldman Sachs kept its Buy rating on Alphabet despite a GAAP EPS miss of $2.12. The bank says the miss was mostly caused by an $8.7 billion investment write-down. Google Cloud revenue grew 103%, and Goldman notes a $561 billion backlog expected to generate significant revenue. Goldman lowered its price target to $385 but still sees Alphabet as a top AI pick.
AI a must-own trade but single-stock bets carry risk says CIO
Bhaskar Laxminarayan, Julius Baer CIO, says AI remains a strong investment because earnings upgrades support AI valuations. The sector still has room to grow. However, betting on single stocks is risky because long-term winners are still unclear. China's growing innovation gap has not yet changed how markets price U.S. tech companies.
Indian AI firm seeks $3 billion debt for Nvidia next-gen chips
AM Intelligence, a Hyderabad-based Indian AI infrastructure company, plans to raise $3 billion in debt to buy Nvidia's next-generation Rubin GPUs. The company plans to invest $6 billion to build 100 MW of computing power split across centers in Malaysia and India. It is among the largest committed buyers of Nvidia's Rubin chips outside North America. Over the next 15 months, the company plans to add 300 MW of capacity with more than $20 billion in investment. Within two years, AM Intelligence aims to build 400 MW of compute capacity using 29,000 Nvidia Rubin chips.
UK announces £600m AI investment plan for innovation and security
The UK Government announced a £600 million AI investment plan on 24 October 2026 in the Autumn Budget. The plan aims to boost economic growth, improve public services, and reduce reliance on overseas technology. Key parts include £165 million for a National AI Biosecurity Initiative with a new AI Biosecurity Lab, £200 million in AI model credits from OpenAI and Anthropic, and partnerships with Google DeepMind. Another £120 million strengthens the UK AI Security Institute, £60 million funds a National Cyber Shield, and £18 million creates a UK Advanced Chip Design Academy. Local AI partnerships in Manchester and Edinburgh will also receive funding.
SafeBreach uses AI to test and beat AI-powered cyber attacks
SafeBreach's CTEM Platform uses purpose-built AI to help organizations defend against AI-powered cyber threats. One in four malicious breaches is now AI-enabled according to the IBM 2026 Cost of a Data Breach Report. In June 2026, Five Eyes cybersecurity agencies warned that AI shortens the time between vulnerability discovery and exploitation. SafeBreach offers three core capabilities: using AI to automate validation workflows, testing AI applications against attacks mapped to OWASP Top 10 for LLM Applications and MITRE ATLAS, and validating defenses against AI-generated malware and spyware.
New Reinforced Adversarial Learning method trains reasoning LLMs from reference samples
Researchers developed Reinforced Adversarial Learning (RAL), a method that uses a co-trained LLM judge to turn reference answers into rewards for post-training reasoning models. RAL was tested on integer multiplication, code-based arithmetic-expression synthesis, and SMILES-to-formula translation, matching direct RLVR results. The method was also applied to scientific abstract generation for U.S. NIH grant proposals. Human scientists and LLM judges found RAL outputs harder to distinguish from real abstracts than frontier-LLM baselines. This approach could help train reasoning models in scientific and other high-value domains.
Stock market reacts to concerns over AI spending
The New York Stock Exchange big board was watched live as the market reacted to concerns about the current artificial intelligence boom. Investors are worried about how much money is being spent on AI. The live broadcast showed real-time market responses to these worries.
New founder archetypes emerging in the AI age
A new wave of founder archetypes is emerging in the AI age. The AI Pioneer creates new industries like Alphabet's DeepMind does in healthcare and finance. The AI Enabler builds platforms like NVIDIA does for other AI businesses. The AI Disruptor challenges existing industries like Uber does with transportation. The AI Innovator transforms current industries like Tesla does with electric vehicles.
Making deliberate decisions for enterprise AI value
A report based on DataIQ Guild 2026 discussions addresses how leaders should handle enterprise AI growth. Five leadership disciplines were identified including confidence grounded in reality and enterprise choices over demand-led reactions. Databricks identified four foundations for navigating AI decisions: context, control, cost and choice. The report says success requires informed decisions and preserving choice rather than waiting for certainty.
Sources
- Microsoft adds security tools for AI agents in 365
- Akeneo Expands AI Agents Across Product Data Operations With New Enterprise Controls
- SentinelOne Launches Agentic Detection and Response Offering for Enterprise AI Security
- Google Just Burned $9 Billion in a Quarter. Is the AI Spending Party Getting Dangerous?
- Why Goldman Sachs Still Calls Alphabet a Top AI Pick Despite the Earnings Miss
- AI remains a must-own trade, but single-stock bets are risky: CIO
- Indian AI infrastructure firm reportedly seeks to raise $3 billion of debt to buy Nvidia's next-generation chips
- UK chancellor sets out £600m AI investment plan to boost British innovation and security
- Use AI, Test AI, Beat AI: How SafeBreach’s Purpose-Built AI Capabilities Power a Proactive Approach Designed to Combat the New Threat Landscape
- Learning to Reason from Reference Samples Alone: Reinforced Adversarial Learning for Modern LLM Post-Training
- Watch live: Stock market reacts amid concerns about artificial intelligence spending
- New Founder Archetypes in the AI Age
- Deliberate decisions and enterprise AI value
Comments
Please log in to post a comment.