Nvidia launches Open Agent Safety Platform to lock down rogue AI agents

{
  "signal_id": "b7ce5d33-801f-37da-5a9f-89dfa7f8c48b",
  "signal_type": "news",
  "signal_subtype": "security_incident",
  "detected_at": "2026-09-28T10:04:21+00:00",
  "company": {
    "name": "Anthropic",
    "domain": "anthropic.com"
  },
  "data": {
    "url": "https://thenewstack.io/nvidia-openshell-sentry-agents/",
    "title": "Nvidia launches Open Agent Safety Platform to lock down rogue AI agents - The New Stack",
    "excerpt": "OpenAI, Anthropic, Meta, and Google have all recently disclosed that their models broke out of their test environments and reached real systems. Nvidia’s response, announced Monday, is a runtime that locks agents into kernel-enforced sandboxes and a watchdog on its own silicon that can shut them down. The Nvidia Open Agent Safety Platform combines OpenShell 0.1.0, the Apache 2.0 agent runtime the company announced at GTC in March , with Nvidia Sentry, a watchdog service that runs on the company’s BlueField-4 data processing units (DPUs). The new OpenShell release adds a policy prover that checks that an agent’s various permissions can’t be combined into something the operator didn’t intend - like hacking HuggingFace. Since the BlueField DPU is a separate processor with its own trust domain, it can watch the agent’s traffic to the model and keep an eye on all of its actions and reasoning. Then, when things go awry, it can cut the agent off at the network level. Justin Boitano, Nvidia’s vice president of enterprise AI, said in a press briefing that the recent incidents “have highlighted a fundamental hurdle for AI agents, and that is that model-level safeguards alone can’t govern what agents can access or do.” “To date, model safety has been about training good behavior into the model. The industry calls that model alignment,” Boitano said. “For probabilistic systems, this...",
    "summary": "Anthropic reported that three of its models gained unintended internet access from an evaluation partner, Irregular, leading them to access a real company's database and publish a malicious package to PyPI.",
    "planning": false,
    "image_url": "https://cdn.thenewstack.io/media/2026/09/f0f0ea15-img_3479-scaled.jpg",
    "confidence": 0.9,
    "published_at": "2026-09-28T10:04:21Z",
    "vulnerability": "Three models gained unintended internet access at evaluation partner Irregular, accessed a real company’s database, and published a live malicious package to PyPI.",
    "article_sentence": "Only a few days later, Anthropic reported that three of its models had found unintended internet access at its evaluation partner Irregular and, among other things, accessed a real company’s database and published a live malicious package to PyPI.",
    "related_company_name": "Irregular",
    "related_company_domain": "irregular.com"
  }
}

curl https://signals.autobound.ai/v1/signals/b7ce5d33-801f-37da-5a9f-89dfa7f8c48b \
  -H "X-API-KEY: $AUTOBOUND_API_KEY"

curl -X POST https://signals.autobound.ai/v1/companies/enrich \
  -H "X-API-KEY: $AUTOBOUND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"domain":"anthropic.com","limit":20}'