The solution to the AI safety crisis is more AI.

Explore on Amazon

Trump-Xi summit and calls for global standards spark debate on frontier risk in international business

The massive scale of security incidents points to control problems for AI companies.

 NVIDIA has introduced the Open Agent Safety Platform as a security architecture for controlling autonomous AI agents from testing through deployment. Announced on September 28, 2026, the architecture combines open-source runtime controls with hardware-based monitoring, placing security boundaries outside the AI model itself. This design recognizes that prompt-level safeguards alone may not prevent an agent from accessing unauthorized files, tools, networks or services. Instead, NVIDIA’s approach connects agent permissions to the wider software, compute and hardware stack.  The software foundation is NVIDIA OpenShell, a secure runtime that places each AI agent inside an isolated execution environment. It can define and enforce rules governing filesystem access, processes, credentials, network connections, APIs and external tools. Operators can convert instructions into verifiable policies before an agent begins work, while OpenShell traces actions and records policy decisions in an audit trail. Because these restrictions operate outside the model and agent framework, they can apply to both open and closed AI models.  OpenShell is designed to act as the first enforcement layer in the architecture. For example, an enterprise agent authorized to retrieve an invoice from one folder could be blocked from opening unrelated files, modifying records or connecting to unapproved services. NVIDIA says the software runs with minimal overhead on its Vera CPUs, while its open-source design can be extended to third-party computing platforms, including Arm and Intel systems. This makes the runtime layer more portable than a security system tied entirely to one model or application. The second major layer is NVIDIA Sentry, an out-of-band watchdog included in the reference system design. Running on NVIDIA BlueField-4 data processing units, Sentry monitors agent behaviour independently of the agent’s operating environment. Through NVIDIA’s DOCA software, it can inspect requests and responses, verify identities and enforce access policies covering data, tools, APIs and services. If an agent attempts to cross its permitted boundary, Sentry is designed to quarantine and stop it in milliseconds, creating a hardware-backed response when software controls are bypassed.  Together, OpenShell and Sentry form a layered AI-agent security architecture rather than a standalone product review. OpenShell governs what an agent is allowed to do, while Sentry provides independent monitoring and containment below the software layer. The broader model gives developers a way to combine policy verification, runtime isolation, continuous telemetry and hardware enforcement across agent deployments. NVIDIA has made OpenShell and related skills available through its developer resources and GitHub, allowing organisations to examine and adapt the architecture as autonomous systems move into production.

Installing Meta's new Muse AI assistant on your MacBook could open the door to cybercriminals. Security researcher Patrick Wardle discovered a zero-day vulne...

AI is eroding critical institutions—journalism, education, courts, and political processes—while strengthening authoritarian repression.

A canceled model release, simulated agent attacks, government action and investor warnings made for one extraordinary day in AI security.

Similar performance, security tradeoffs also seen in current public models.

Nvidia's security push, unresolved liability, China's chip IPO, and why workplaces aren't ready.

Autonomy and AI are becoming increasingly important elements of today's military forces. We're seeing the likes of drones, turrets, and even vehicles such as aircraft and boats...

2026年9月29日 メタの自律型AIエージェントMuse、利用者に無断で自宅住所を買い手に送信した一件。 トロント在住のマット・ロブ氏はFacebook Marketplaceの出品対応を、メタの新AIエージェント「Muse」（利用者に代わって作業を代行するAI）に任せていた。ロブ氏によると、Museは購入希望者のウスマン氏に本人の承認なく自宅住所を伝え、本人を装って「家にいる」と返信した。...

The breach included Social Security numbers and details about jobs people held.

Eleven former senators and senior defense, intelligence, and energy officials will help identify high-impact AI applications for US security.

YouTube video by Pivot to AI

Gemini 3.8 Flash Cyber and Wiz's Red Agent team up to protect hospitals, public transit, and tech

Praeses has secured a Phase III SBIR contract from the GSA to enhance AI-driven data fusion solutions across U.S. federal agencies, marking a significant advance in national security technology.

Nvidia introduces software and hardware controls for AI agent safety. Learn how the Open Agent Safety Platform may strengthen your AI security posture and compare built-in vs hardware safeguards.

Acting Home Affairs Minister Richard Marles says government systems must "keep up" with the rapid changes sparked by artificial intelligence. #cyber #security #ai #federal #government

Azure cloud environments are facing a new and destructive threat from the JadePuffer ransomware operator. The group uses AI agents to automate reconnaissance, steal credentials, and destroy critical cloud resources, all with minimal human intervention once the attack begins. The malware emerged in July and evolved quickly. Cloud security firm Sysdig flagged its use of […]