Tech companies are teaming up to address security concerns around AI agents. Okta, AWS, Google Cloud, CrowdStrike and nine other vendors formed the Blueprint Alliance, a coalition working on an open, multi-vendor security framework. The goal is to help organizations track where agents operate, control what they access and contain them when things go wrong. GE Appliances and World Central Kitchen will advise on real-world testing of the architecture.
At its Oktane conference, Okta also launched its generally available Okta for AI Agents platform. The company made Agent SSO free across its products so agents can access systems without repeated logins. New tools include Agent Gateway, which logs tool calls and blocks unauthorized actions, and Shadow AI Agent Discovery for Endpoints, which finds unmanaged agents on employee devices. Okta cited deployments at Ramp, Yahoo and Dell Technologies. The alliance now has 12 vendors, as Gartner predicts over 150,000 enterprise agents by 2028.
Nvidia CEO Jensen Huang pushed back on AI doomsday predictions while speaking at Salesforce's Dreamforce conference in San Francisco on September 15. He called predictions about AI doomsday scenarios hurtful and addressed growing public concerns about AI's future impact. Separately, Huang has close access to President Trump, who is set to meet with Chinese leader Xi Jinping with AI on the agenda. Huang's perspective on the matter could influence upcoming policy decisions.
Meta is taking a different approach to AI agents than competitors like OpenAI and Anthropic. Rather than focusing on raw model performance, Meta is building an agent that leverages user data to anticipate what people want to buy. This strategy lets Meta cut out middlemen and position itself directly as a marketplace. The company is prioritizing user data insights over building the best AI model.
Nvidia and Aolani agreed to deploy 22,000 GPUs across new AI factory sites in Southeast Asia. The partners plan over 100 megawatts of new AI factory capacity with major hubs in Malaysia and the Philippines. The buildout aims to position the region as a global node for AI compute and data center infrastructure.
Other developments include Accenture investing in Within through Accenture Ventures. Within is an AI platform that helps organizations map and automate business processes. Palantir CEO Alex Karp questioned whether frontier AI developers can handle the cost of major legal claims, with Palantir shares rising about 4.1% to $192.455. The company's quarterly filing warns that generative or agentic AI problems could bring liability and damage its reputation. Cofense also expanded its AI-driven Phishing Defense Platform with a Competency Dashboard that measures how employees handle phishing threats, with new data connections planned for Q4 2026.
Friday Harbor, an AI pre-underwriting platform, expanded support for investor loans including debt service coverage ratio loans. The tool helps loan officers build complete loan files in real time. Separately, a venture capitalist was banned from restaurant reservation platform Resy for using a personal AI assistant to make dinner reservations, sparking debate about AI assistants in everyday settings.
Key Takeaways
- <ul><li>Okta, AWS, Google Cloud, CrowdStrike and nine other companies formed the Blueprint Alliance to create a shared security framework for AI agents</li><li>Okta launched its generally available Okta for AI Agents platform with free Agent SSO, Agent Gateway and Shadow AI Agent Discovery for Endpoints</li><li>The Blueprint Alliance now includes 12 vendors, with Gartner predicting over 150,000 enterprise agents by 2028</li><li>Nvidia CEO Jensen Huang spoke at Salesforce's Dreamforce on September 15, pushing back on AI doomsday predictions as hurtful</li><li>Meta is focusing on user data for its new AI agent to anticipate purchases rather than competing on raw model performance against OpenAI and Anthropic</li><li>Nvidia and Aolani plan to deploy 22,000 GPUs and over 100 megawatts of AI factory capacity in Malaysia and the Philippines</li><li>Accenture invested in Within through Accenture Ventures to help enterprises map and automate business processes with AI</li><li>Palantir CEO Alex Karp questioned whether frontier AI developers can handle major legal claim costs, with shares rising about 4.1% to $192.455</li><li>Cofense expanded its Phishing Defense Platform with a Competency Dashboard, adding phishing remediation data to Command Center in Q4 2026</li><li>A venture capitalist was banned from Resy for using an AI assistant to make dinner reservations, highlighting tensions over AI bot usage</li></ul>
Tech companies form Blueprint Alliance to secure AI agents
Okta, AWS, Google Cloud, CrowdStrike and nine other companies formed the Blueprint Alliance to create a shared security framework for AI agents. The framework helps organizations track where agents operate, control what they access and contain them if problems arise. GE Appliances and World Central Kitchen will advise on testing the architecture in real-world settings. The group aims to build an open, multi-vendor standard rather than a single-company product.
Okta launches AI agent security platform at Oktane event
Okta introduced its generally available Okta for AI Agents platform and the Blueprint Alliance at its Oktane conference. The company said Agent SSO is now free across its products, allowing AI agents to access systems without repeated logins. Okta also released new tools including Agent Gateway, Shadow AI Agent Discovery for Endpoints and support for third-party identity providers. The company cited deployments at Ramp, Yahoo and Dell Technologies.
Okta adds new tools to rein in AI agent access
Okta unveiled runtime enforcement features and a broader kill switch for its AI agent security platform at Oktane. Agent Gateway serves as a centralized control point that logs tool calls and blocks unauthorized actions. Shadow AI Agent Discovery for Endpoints identifies unmanaged agents on employee devices. The Blueprint Alliance now includes 12 vendors working toward an open standard for AI agent security as Gartner predicts over 150,000 enterprise agents by 2028.
Okta CEO discusses new AI agent security tools
Okta CEO and co-founder Todd McKinnon appeared on Squawk on the Street to discuss new tools for managing AI agents. He spoke about the Blueprint Alliance coalition, which aims to help companies address AI security concerns. McKinnon emphasized the balance between innovation and security in enterprise AI adoption.
Jensen Huang pushes back on AI doom predictions
Nvidia CEO Jensen Huang spoke at Salesforce's Dreamforce conference in San Francisco on September 15. He said predictions about AI doomsday scenarios are hurtful. Huang addressed growing public concerns about the future impact of artificial intelligence.
Jensen Huang shares AI views with President Trump
Jensen Huang is the CEO of Nvidia, a leading AI company. He has close access to President Trump, who is set to meet with Chinese leader Xi Jinping. AI is one of the topics on the agenda. The article explores how Huang's perspective on AI might influence Trump's decisions on the matter.
Meta uses user data to power new AI agent
Meta is focusing on user data rather than building the best AI model. Its new AI agent knows what users want to buy before they do. This allows Meta to cut out middlemen and become the marketplace itself. The approach differs from competitors like OpenAI and Anthropic that focus on raw model performance.
RankCert certifies safe AI tutor selection decisions
RankCert is a system that helps determine when simulated learners can safely select an AI tutor. It certifies a tutoring policy only when multiple evaluation methods agree on the same choice. The system was tested in 1,280 settings across five families of scenarios. RankCert reduced decision loss compared to full-coverage methods but only certified choices in stable situations.
Friday Harbor adds AI loan support for investors
Friday Harbor is an AI pre-underwriting platform that helps loan officers build complete loan files in real time. The company announced expanded support for investor loans. This includes debt service coverage ratio loans. The tool aims to help officers assemble compliant loan files more efficiently.
Accenture invests in Within to boost enterprise AI
Accenture made an investment in Within through Accenture Ventures. Within is an AI platform that helps organizations map and automate business processes. The partnership aims to help clients accelerate AI adoption across their enterprises. The deal combines strategic investment with a new partnership between the two companies.
Venture capitalist banned from Resy for using AI assistant
A venture capitalist was banned from the restaurant reservation platform Resy for using a personal AI assistant to make dinner reservations. Resy has a strict policy against bots to prevent scalpers and ensure reservations are made by real people. The incident has started a debate about whether AI assistants help or harm the dining experience for both diners and restaurants.
Palantir CEO makes AI liability a key product focus
Palantir CEO Alex Karp questioned whether frontier AI developers can handle the cost of major legal claims. Shares of Palantir rose about 4.1% to $192.455 on Wednesday. The company's quarterly filing warns that generative or agentic AI problems could bring liability and damage its reputation.
Cofense expands AI phishing defense with new Competency Dashboard
Cofense expanded its AI-driven Phishing Defense Platform through Cofense Command Center. The new Competency Dashboard measures how employees recognize, report and respond to phishing threats. Phishing Remediation data from Cofense Triage and Vision will connect to Command Center in Q4 2026, adding real-world threat signals to training and behavior data.
Nvidia and Aolani to deploy 22,000 GPUs in Southeast Asia
Nvidia and Aolani agreed to deploy 22,000 GPUs across new AI factory sites in Southeast Asia. The partners plan over 100 megawatts of new AI factory capacity with major hubs in Malaysia and the Philippines. The build out aims to position Southeast Asia as a global node for AI compute and data center infrastructure.
Sources
- Blueprint Alliance targets shared security architecture for AI agents
- Okta Unveils AI Agent Security Blueprint Alliance at Oktane
- Okta enhances AI agent security with new platform features and Blueprint Alliance
- Okta CEO Todd McKinnon on new tools for managing AI agents: We want innovation and security
- "Enough predictions": Jensen Huang unloads on AI doomsday fears
- Jensen Huang has President Trump’s ear. What might he be telling him about AI?
- Meta doesn't need the best AI model — it has something better: you. While OpenAI and Anthropic compete on raw model performance, Meta's new agent wins on data. They know what you want to buy before you do, and now they can finally act on it, cutting out the middlemen and becoming the marketplace itself. Would you actually use an AI agent that knows you this well? Watch the full episode, In The Loop, on Youtube.
- RankCert: When Can Simulated Learners Safely Select an AI Tutor? Robust Decision Certification Under Structural Uncertainty
- Friday Harbor adds AI-pre-underwriting support for investor and DSCR loans
- Accenture and Within Help Clients Accelerate AI Across the Enterprise Through Strategic Investment and New Partnership
- Now AI is trying to gobble up dinner reservations
- Palantir Jumps Over 4% as Karp Makes AI Liability the Product Test
- Cofense Expands AI-Driven Phishing Defense Platform to Advance Secure Behavior Management
- Nvidia (NVDA) Targets 22,000 GPUs For New Southeast Asia AI Factory Hubs
Comments
Please log in to post a comment.