English 
SentinelOne Japan Co., Ltd., the Japanese subsidiary of SentinelOne®, a global leader in autonomous cybersecurity, has announced a new lineup of AI security products designed to give defenders a decisive advantage. These new products further strengthen SentinelOne’s market-leading AI security portfolio by covering both security that protects AI itself from threats and the use of AI to automate and transform security operations. All new products, from protecting autonomous agents to executing complete agent-based (autonomous) investigations with a single click, were showcased at RSA Conference 2026.
As organizations accelerate the adoption of AI to drive innovation, expand operations, and improve productivity, AI itself is becoming a new attack surface and a major source of security risk. Gartner, one of the world’s largest IT research and advisory firms, predicts that “AI cybersecurity,” encompassing both security that “protects AI” and security that “enhances AI,” will be one of the most important and fastest-growing markets in AI-related spending over the next few years. In its January 2026 forecast, spending on AI cybersecurity is expected to grow at an astonishing compound annual growth rate (CAGR) of 73.9% from 2024 to 2029, more than double the growth rate of overall AI-related spending.
プロンプトAIエージェントセキュリティの新リリース
Built upon SentinelOne’s comprehensive end-to-end AI protection approach (see related blog post ) , Prompt AI Agent Security provides a new control plane for real-time detection and governance of AI agents and agent-based workflows. Leveraging SentinelOne’s Autonomous Security Intelligence across endpoints, the cloud, and identity, it extends its unique AI and automation to the agentic layer. This enables real-time monitoring, control, and policy enforcement of agent-to-agent interactions at machine speed. As a result, it provides complete threat landscape visibility, risk assessment, and policy enforcement for all MCP servers operating in the customer’s system environment. In preview, customers can centrally manage the posture (security state) of all AI agents and agent-based workflows, automatically remediating agent-based behavior before malicious actions occur, such as an OpenClaw agent sending corporate data to an external endpoint without user knowledge, or a Claude Cowork agent attempting to escalate privileges across the organization’s systems through a series of unauthorized operations.
こちらもお読みください： Fastlyとリーガ・エスパニョーラがコンテンツ違法コピー撲滅のために協力
ニューリリースプロンプトAIレッドチーム
Prompt AI Red Teamingは、セキュリティチームと製品チームに、自社で開発・提供するAIアプリケーションのための画期的なテストとハードニング機能を提供します。開発者が企業環境で新しいツール、アプリケーション、ワークフローを構築するためにAIエージェントを採用するにつれ、従来のセキュリティテスト手法ではAI固有の脅威に対処することができなくなっています。Prompt AI Red Teamingを使用することで、企業はビジネスや顧客を重大なリスクにさらすことなく、実際のAI攻撃（プロンプトインジェクション、ジェイルブレイク、特権の昇格、データポイズニングなど）をシミュレートし、リリース前にAIアプリケーションを強化することができます。さらに、モデルや脅威の進化に伴うリスク（モデルのドリフト、新しい脆弱性の検出、新しい攻撃ベクトルなど）を継続的に評価することができます。.
Purple AI’s automated research feature is now available to the general public.
At RSA Conference 2026, SentinelOne further advanced Purple AI, making it generally available to perform automated investigations with a single click. This new feature, natively integrated into the Singularity™ platform, allows analysts to conduct fully autonomous investigations with a single click. Going beyond traditional, stylized playbooks, Purple AI autonomously collects evidence of compromise across the stack and integrates threat data to build a complete attack timeline in real time. It then provides evidence-based and explainable verdicts, instantly triggering closed-loop automated remediation via Singularity Hyperautomation. These processes are executed within a rigorous governance mechanism that maintains analyst involvement.
Purple AIは、エージェントベースのフレームワークと人間レベルの推論機能を活用し、高度なサイバー攻撃を阻止するために必要なスピード、スケール、スキルをセキュリティチームに提供します。また、直感的なヒューマン・イン・ザ・ループの自動化により、人間の防御担当者を強化し、退屈な作業から解放して、より戦略的な対応に集中できるようにします。.
RSA Conference 2023で初公開、, センチネルワン‘s Purple AI has been proven and utilized in thousands of real-world SOCs (Security Operations Centers) and customer environments, establishing itself as the definitive agent-based AI security analyst in the market. It is also currently one of the most widely deployed products. In its Q4 2026 earnings announcement, SentinelOne reported that Purple AI sales achieved a record-high attachment rate, with Purple AI accounting for more than 50% of the licenses sold during that quarter.
Agentic Auto Investigation integrates Purple AI’s inference capabilities into the most challenging parts of security operations, enabling machine-speed, full-depth, multi-source deep forensic investigations without the need for additional data routing or extended permissions. All of these features are delivered within the fully regulatory-compliant Singularity data platform and AI SIEM.
As a result, Purple AI’s new agent-based automated investigation function completes security investigations that previously took hours or days in seconds or minutes, creating an environment where defenders can fight on equal footing with AI-driven, machine-speed attacks.
Purple AI’s Auto Investigations feature is available to all Purple AI Analyst customers without any additional installation or configuration.
ソース PRタイムズ
日本語