Back to Research Hub

Daily Enterprise Tech Brief | 23 September 2026

23 September 2026: Today's top AI releases. Understand use cases and business value from Anthropic, Palo Alto Networks, OpenAI, Google Cloud and Salesforce

23 September 2026Primary-source referencesReading time: 6 minutes

Executive answer

POC focus: Teams can test complex coding and research workflows across existing cloud estates while managing cost, retention, and cybersecurity access boundaries.

Claude Opus 5.5 Goes Multi-Cloud

What's new: Anthropic launched Claude Opus 5.5 across its API, AWS, Google Cloud, and Microsoft Azure, with lower token and cache-read pricing.

What it means for my business: Teams can test complex coding and research workflows across existing cloud estates while managing cost, retention, and cybersecurity access boundaries.

Anthropic: AnthropicSource date: 2026-09-22

Unit 42 Launches Continuous AI Defense

What's new: Palo Alto Networks launched Unit 42 Continuous Frontier AI Defense, a worldwide subscription service for continuous testing, exploit validation, and remediation prioritization.

What it means for my business: Security teams can focus remediation on verified attack paths across applications, identities, cloud assets, code, APIs, and networks rather than unvalidated alerts.

Palo Alto Networks: Palo Alto NetworksSource date: 2026-09-22

OpenAI Releases GPT-6 Sol and Luna

What's new: OpenAI released GPT-6 Sol and GPT-6 Luna in the Responses and Chat Completions APIs, with text and image inputs.

What it means for my business: Route complex coding and agent work to Sol while assigning high-volume extraction, classification, and routing tasks to lower-cost Luna.

OpenAI: OpenAISource date: 2026-09-22

Google Adds D&B Risk Analytics

What's new: Google Cloud made the D&B Risk Analytics data store generally available in Gemini Enterprise for KYB, due diligence, screening, and supplier-risk workflows.

What it means for my business: Risk and procurement teams can gather licensed counterparty evidence through natural-language workflows while retaining formal review and approval controls.

Google Cloud: Google CloudSource date: 2026-09-22

Telstra Puts Agentforce Into Production

What's new: Telstra deployed Salesforce Agentforce for refund and interim-service workflows, plus monitoring of Salesforce access, configurations, and permissions for risky patterns.

What it means for my business: High-volume service operations can automate deterministic customer steps while directing representatives toward exceptions and surfacing SaaS permission risks sooner.

Salesforce and Telstra: Salesforce and TelstraSource date: 2026-09-22

Alibaba Cloud Expands Agent Governance

What's new: Alibaba Cloud announced AgentCore, Agent Security Center, and Agent Context for enterprise agent lifecycle management, compliance, real-time context, and long-term memory.

What it means for my business: Organizations standardizing on Alibaba Cloud can centralize agent controls, security operations, and business context instead of assembling separate lifecycle components.

Alibaba Cloud: Alibaba CloudSource date: 2026-09-22

VAST Previews Confidential AI DataEnclave

What's new: VAST Data previewed DataEnclave, using NVIDIA Confidential Computing to run AI models against sensitive data on customer or trusted cloud infrastructure.

What it means for my business: Regulated and sovereign environments can evaluate hardware-attested AI execution without exposing sensitive records or model intellectual property to infrastructure operators.

VAST Data: VAST DataSource date: 2026-09-22

Darktrace SECURE AI Reaches General Availability

What's new: Darktrace made SECURE AI generally available, adding AI-use discovery, behavioral policy assessment, prompt analysis, and shadow-AI detection across major providers.

What it means for my business: Security teams can investigate risky prompts and agent behavior across multiple AI services, then enforce policy or accelerate incident response.

Darktrace: DarktraceSource date: 2026-09-22

GitHub Copilot Adds Claude Opus 5.5

What's new: GitHub Copilot added Claude Opus 5.5 for eligible Business and Enterprise plans across IDEs, coding agent, CLI, web, and mobile surfaces.

What it means for my business: Engineering leaders can govern access to a higher-capability coding model through model policy while monitoring usage-based provider pricing.

GitHub: GitHubSource date: 2026-09-22

Proofpoint Announces Semantic Agent Policy Controls

What's new: Proofpoint announced Semantic Business Policies and Agentic Insights for translating governance requirements into runtime controls and analyzing AI interactions and tool use.

What it means for my business: Governance teams could adapt controls through business-language policy definitions while identifying risky agent behavior and policy gaps across AI workflows.

Proofpoint: ProofpointSource date: 2026-09-22

Salt Links AI Attacks to Tool Activity

What's new: Salt Security added native AI Detection and Response for prompt injection, jailbreaks, unsafe model behavior, and downstream MCP, tool, and API activity.

What it means for my business: Application-security teams can investigate connected attack sequences, improving containment decisions when apparently valid agent tool calls produce unsafe business actions.

Salt Security: Salt SecuritySource date: 2026-09-22

GitHub Sets Post-Quantum SSH Transition

What's new: GitHub announced stronger RSA key requirements, removal of older SSH algorithms, and hybrid ML-KEM post-quantum key exchange for specified services.

What it means for my business: Platform teams can inventory clients, CI tools, automation libraries, and keys before enforcement disrupts developer access or build pipelines.

GitHub: GitHubSource date: 2026-09-22

Harness Secures AgentCore Gateway Tool Calls

What's new: Harness announced an Amazon Bedrock AgentCore Gateway integration for AI posture management, AI firewall controls, continuous discovery, and trace-level security context.

What it means for my business: AWS security teams can reconstruct unsafe sequences across prompts, MCP servers, tools, data access, and downstream actions with less manual log correlation.

Harness: HarnessSource date: 2026-09-22

DigitalOcean Previews Managed Agents

What's new: DigitalOcean launched Managed Agents in public preview, combining isolated runtimes, governed MCP tool access, session persistence, and serverless inference.

What it means for my business: Engineering teams can deploy agentic workflows without independently assembling execution, credential brokering, tool governance, and inference layers.

DigitalOcean: DigitalOceanSource date: 2026-09-22

Accenture and AWS Launch Six Offers

What's new: Accenture Edge and AWS made six packaged cloud, AI, security, migration, optimization, customer experience, and red-teaming offerings generally available through AWS Marketplace.

What it means for my business: Mid-market IT teams can procure defined modernization services through an existing marketplace channel, reducing partner-selection and project-initiation friction.

Accenture and AWS: Accenture and AWSSource date: 2026-09-22

Practical questions

What changed in Claude Opus 5.5 Goes Multi-Cloud?

Anthropic launched Claude Opus 5.5 across its API, AWS, Google Cloud, and Microsoft Azure, with lower token and cache-read pricing.

What business use case does Claude Opus 5.5 Goes Multi-Cloud support?

Teams can test complex coding and research workflows across existing cloud estates while managing cost, retention, and cybersecurity access boundaries.

What changed in Unit 42 Launches Continuous AI Defense?

Palo Alto Networks launched Unit 42 Continuous Frontier AI Defense, a worldwide subscription service for continuous testing, exploit validation, and remediation prioritization.

What business use case does Unit 42 Launches Continuous AI Defense support?

Security teams can focus remediation on verified attack paths across applications, identities, cloud assets, code, APIs, and networks rather than unvalidated alerts.

What changed in OpenAI Releases GPT-6 Sol and Luna?

OpenAI released GPT-6 Sol and GPT-6 Luna in the Responses and Chat Completions APIs, with text and image inputs.

What business use case does OpenAI Releases GPT-6 Sol and Luna support?

Route complex coding and agent work to Sol while assigning high-volume extraction, classification, and routing tasks to lower-cost Luna.

Related research

Keep building the complete picture

Move from signal to proof

QualifiedPOC.ai helps you replace months of repetitive vendor evaluations with one deep, self-paced conversation. Our vendor-agnostic experts handle the rest, all the way to a successful POC outcome.

Start live chat with an AI expert

Enjoying our research hub feeds?

Get them daily morning on email.

Check eligibility for free email subscription