Microsoft is expanding its AI-led security operations with Perception, now generally available alongside core Integrated Security Operations Center capabilities in Defender. Perception connects security signals and customer information to agents that investigate incidents and can take limited actions, including containing devices or revoking access. Security teams set the rules for automated playbooks, and high-impact actions require human approval. Defender agents also support alert triage and attack investigations; controls can govern discovered AI agents. Microsoft says its ecosystem processes more than 165 trillion security signals daily. Signals from AWS, Okta and Salesforce, as well as application penetration testing and cross-cloud remediation tools, remain in preview.
Other security vendors are targeting risks and fixes across AI and software. Wiz’s AI Security Agent entered public preview on Nov. 16, 2026, examining AI applications, agents, code and cloud permissions. Wiz says it found up to five times as many agents as traditional discovery engines, but has not disclosed its testing method or false-positive rate. Semgrep introduced its Agentic AppSec Platform, connecting security analysis with AI coding tools through Semgrep Guardian and an MCP server. Its internal benchmarks found more than twice as many genuine vulnerabilities using its agents and Mandoline engine than frontier models alone; Agentic Security Workflows cut token costs by an average of 65%. The platform supports bring-your-own-key models, dedicated hosting and an EU multi-tenant region.
AI’s promise also faces evidence and data-quality constraints. A review of 70 systematic reviews on AI in diabetes, covering research from 2014 through October 2025, found rapid growth but uneven evidence across diabetes types and complications. Its authors say many algorithms remain far from routine clinical use because of research-quality and implementation gaps. Separately, researchers propose Grounded Synthetic Intelligence: using synthetic data while tying it to real observations. They argue it could help when sensitive real-world data are limited, but caution against using synthetic data without real data and careful validation, and call for stronger methods and standards.
In chip design, OpenAI says AI helped AMD achieve gains of up to 14% on flagship GPU intellectual property. Synopsys plans to add OpenAI intelligence to electronic design automation workflows, according to EE Times. The report cautions that AI-generated chip cores can omit counters when built from specifications, making tighter area and timing checks—and testing model rewrites—important before production. Meanwhile, Ed Yardeni says the AI trade remains “alive and well” after Nvidia’s latest earnings. He expects computing demand and AI investment to grow, while semiconductor and infrastructure limits, tighter financing and opposition to data centers may temper spending and reduce bubble risk.
Companies are also applying AI to business workflows and infrastructure. Tacton announced on Nov. 17, 2026, an expansion of its generative AI sales assistant for complex manufacturing quotes, saying it can improve quote accuracy by 25% and reduce preparation time by 30%. Accenture outlines ways to lower token costs on Amazon Bedrock, including prompt and semantic caching, batch inference, shorter agent context and document compression. Across four client engagements, reported reductions ranged from about 40% to 85%; a telecom operator cut costs by 68% on $12 million in annual token spending. Blue Origin plans to launch Google’s Project Suncatcher AI server satellites in 2029, the project’s first dedicated heavy-lift mission. Google aims for sun-synchronous orbit and computing capacity roughly equal to a data-center rack; an earlier satellite launched on SpaceX’s Falcon 9.
AI-enabled services and market sentiment round out the developments. Romanian fintech Cosmio launched after raising €200,000 from its founders and an angel investor. Its subscription service starts at RON 290 including VAT, connects to banks through Salt Edge and to Romania’s ANAF Virtual Private Space and e-Factura, and uses CECCAR-certified accountants to review and submit tax returns. Cosmio targets about 500 clients in its first year and is preparing financial-management services; Rubik Hub’s MVP Accelerator selected it, with strategic partner Nucleo Ventures potentially offering a selected startup €150,000. Vibe-Trading, meanwhile, combines research, backtesting, portfolio analysis, paper trading and optional broker execution, recording sources and settings while keeping research separate from live trading through an authority gate. Etsy shares fell 7.3% amid market pressure and investor concern over its AI policy, which the company updated on Oct. 22 to allow member-created content to improve marketplace-supporting AI models. Etsy has warned that negative views of AI could hurt trust and engagement.
Key Takeaways
- Microsoft made Perception and core Defender Integrated Security Operations Center capabilities generally available; high-impact agent actions require human approval, while some integrations and tools remain in preview.
- Wiz says its public-preview AI Security Agent found up to five times as many agents as traditional discovery engines, but has not shared its testing method or false-positive rate.
- Semgrep says its agent-and-Mandoline approach found more than twice as many genuine vulnerabilities as frontier models alone in internal benchmarks, while its workflows reduced token costs by an average of 65%.
- A review of 70 diabetes AI systematic reviews found uneven evidence and significant barriers to routine clinical use; researchers proposing Grounded Synthetic Intelligence call for validation against real observations and stronger standards.
- OpenAI says AI helped AMD achieve gains of up to 14% on flagship GPU intellectual property; the report stresses chip-core checks for area, timing and missing counters.
- Yardeni says AI investment remains strong after Nvidia’s latest earnings, while semiconductor and infrastructure limits, financing constraints and data-center opposition may slow spending.
- Tacton says its AI assistant for manufacturing quotes can improve accuracy by 25% and cut preparation time by 30%.
- Accenture reports Amazon Bedrock token-cost reductions of about 40% to 85% across four client engagements, including a 68% cut for a telecom operator spending $12 million annually.
- Blue Origin plans a 2029 heavy-lift launch for Google’s Project Suncatcher AI server satellites, designed to provide computing capacity roughly equal to a data-center rack.
- Cosmio launched with €200,000 in funding and targets about 500 clients in its first year; Etsy shares fell 7.3% amid AI-policy concerns and broader market pressure.
Microsoft launches Perception to coordinate AI security agents
Microsoft made Perception generally available to commercial customers as the foundation of its agentic security strategy. The platform coordinates AI agents that detect threats, investigate incidents, and help fix risks, using security signals and automated playbooks. Microsoft says its ecosystem processes more than 165 trillion security signals each day. The company also announced the Integrated Security Operations Center in Defender and new tools to govern AI agents, while some capabilities remain in preview.
Wiz AI Security Agent searches for hidden enterprise AI risks
Wiz launched its AI Security Agent in public preview on Nov. 16, 2026. The tool examines AI applications, agents, code, and cloud permissions to find agents that conventional discovery tools may miss. Wiz says it found up to five times as many agents as traditional discovery engines, but it has not shared its testing method or false positive rate. The tool links possible weaknesses to cloud identities, permissions, and sensitive data, and Wiz advises security teams to check findings against their own systems. Future runtime monitoring is not a confirmed feature of the current preview.
Microsoft Perception links security signals with response agents
Microsoft says Perception and core Integrated Security Operations Center capabilities in Defender are generally available, while application penetration testing and cross-cloud remediation tools remain in preview. Perception connects security signals and customer information to agents that can investigate incidents and take limited actions, such as containing devices or revoking access. Security teams set rules for automated playbooks, and Microsoft says high-impact actions require human approval. Defender agents support alert triage and attack investigation, while identity and data controls can govern discovered agents. Signals from AWS, Okta, and Salesforce are available in preview.
Semgrep launches platform for AI-driven application security
Semgrep introduced its Agentic AppSec Platform to help security teams find, review, and fix vulnerabilities across software development. It connects security analysis with AI coding tools through Semgrep Guardian and a Model Context Protocol server. Semgrep says its internal benchmarks found more than twice as many genuine vulnerabilities with its agents and Mandoline engine than with frontier AI models alone, while Agentic Security Workflows cut token costs by an average of 65%. The platform also includes tools for code analysis, continuous integration, and a persistent context layer. It supports deployment options including bring-your-own-key models, dedicated hosting, and an EU multi-tenant region.
Review finds AI diabetes research faces major clinical barriers
A tertiary review led by Sander M. W. Janssen and colleagues examined 70 systematic reviews on artificial intelligence in diabetes, covering studies published from 2014 through October 2025. The researchers found that AI research has grown quickly, but the evidence remains uneven and divided across diabetes types and complications. The review says many algorithms are still far from routine clinical use because of gaps in research quality and real-world implementation. It covered type 1, type 2, and gestational diabetes, along with complications such as kidney disease, foot ulcers, and diabetic retinopathy.
Researchers propose grounded synthetic data for trusted AI
Researchers propose Grounded Synthetic Intelligence as a way to use synthetic data while keeping it tied to real observations. Their position paper in the International Journal of Computational Intelligence Systems says privacy rules and other limits on sensitive data make it harder to build AI, especially in health and personalized services. The authors say synthetic data can improve models when real datasets are small or restricted, but warn against using it without real data and careful checks. They call for better methods to create and assess synthetic data, including standards for its use in sensitive settings.
OpenAI says AI is improving chip design
OpenAI says AI helped AMD achieve gains of up to 14% on flagship GPU intellectual property. At its 2026 Investor Day in New York, Synopsys said it would add OpenAI intelligence to electronic design automation workflows, according to EE Times. The article also says AI generated chip cores can miss counters when built from specifications, so designers need tighter area and timing checks before production. It urges chip builders to test model rewrites.
Tacton adds an AI sales assistant for complex manufacturing quotes
Tacton announced on November 17, 2026, an expansion of its generative AI sales assistant technology for manufacturers. The assistant helps companies prepare quotes for complex products, and Tacton says it can improve quote accuracy by 25% and cut preparation time by 30%. The company plans to keep investing in AI research and development to improve its solutions.
Romanian fintech Cosmio launches after €200,000 investment
Romanian startup Cosmio has launched after receiving €200,000 from its founders and a private angel investor. It combines automation, artificial intelligence, and accounting expertise in a subscription service starting at RON 290, including VAT. Cosmio connects with banks through Salt Edge open banking and with Romania’s ANAF Virtual Private Space and e-Factura system to collect records, while CECCAR-certified accountants review and submit tax returns. The company aims to reach about 500 clients in its first year and is preparing to add financial management services. Cosmio was also selected for Rubik Hub’s MVP Accelerator, whose strategic partner Nucleo Ventures may offer a €150,000 investment to a selected startup.
Ed Yardeni says AI investment remains strong despite market fears
Yardeni Research President Ed Yardeni said Wednesday that the AI trade remains “alive and well” after Nvidia’s latest earnings. He expects investors to be pleasantly surprised by Anthropic’s upcoming initial public offering and says growing concern about AI is creating stock market buying opportunities. Yardeni expects AI investment and computing demand to keep growing, while pointing to limits from semiconductors and infrastructure, tighter financing, and opposition to data centers. He says these “speed bumps” could help prevent a damaging investment bubble, and expects capital spending, consumer demand, and government spending to support the U.S. economy.
Vibe-Trading links trading research to controlled execution
Vibe-Trading is a framework and workspace for creating, testing, and improving financial trading agents. It brings together market research, financial data, strategy generation, backtesting, portfolio analysis, paper trading, multi-agent collaboration, and optional broker execution. The system records each result with its source, time range, settings, and run, helping users verify and reproduce research. An authority gate keeps strategy research separate from live trading.
Blue Origin to launch Google AI server satellites in 2029
Blue Origin plans to launch Google’s Project Suncatcher AI server satellites in 2029. The mission will be the project’s first dedicated heavy-lift launch. Google aims to place satellites in sun-synchronous orbit with computing capacity roughly equal to a data center rack. A Google AI server satellite previously launched on a SpaceX Falcon 9 rocket.
Accenture outlines six ways to cut Amazon Bedrock AI costs
Accenture describes six ways to manage token spending as companies expand their use of AI on Amazon Bedrock. The methods include prompt caching, batch inference, limiting agent context, compressing documents, and semantic caching. Across four client engagements, reported cost reductions ranged from about 40 percent to 85 percent. A large enterprise reached 85 percent, while a telecom operator cut costs by 68 percent on $12 million in annual token spending. Accenture recommends tracking costs by team and workload, then using billing data and model invocation logs to find areas to improve.
Etsy shares fall amid AI policy backlash and market pressure
Etsy shares fell 7.3 percent as investors weighed concerns about the company’s AI policies and pressure across the broader U.S. market. Etsy updated its AI policy on October 22 and later said the changes let it use member-created content to improve AI models that support its marketplace. The company has also warned that negative views of AI could harm trust and engagement. After rising following its third-quarter results and reaching a 52-week high earlier in November, the stock may have been more exposed to a pullback, according to the article’s analysis.
Sources
- Microsoft Perception Brings AI Agents into Security Ops
- Wiz AI Security Agent Launches to Find Hidden Enterprise AI Risks
- Microsoft Perception coordinates security response agents
- Semgrep Introduces Agentic AppSec Platform to Accelerate AI-Driven Code Security
- Seventy Systematic Reviews Reveal Artificial Intelligence in Diabetes Faces Implementation Barriers
- Grounded Synthetic Intelligence Proposed as Foundational Infrastructure for Regulated Artificial Intelligence
- OpenAI says AI now designs better chips
- Tacton Expands Generative AI Innovation with AI Sales Assistant to Accelerate Complex Manufacturing Quotes
- EUR 200,000 investment in Cosmio FinTech, a Romanian startup combining AI, automation and expertise
- Ed Yardeni Says AI Trade Is 'Alive And Well' After Nvidia Earnings — Sees AI ‘Panic’ Creating Buying Opportunities
- Vibe-Trading: Towards Agent-Native Trading
- Blue Origin will launch Google’s AI server satellites in 2029
- Optimize AI economics on Amazon Bedrock with Accenture’s sequenced approach
- Etsy Shares Slide as AI Policy Backlash and Risk-Off Trading Weigh on Sentiment
Comments
Please log in to post a comment.