NVIDIA's Open Agent Safety Platform pairs OpenShell sandboxes with Sentry, a BlueField-4 watchdog that enforces agent safety outside the agent. Over 100 partners are involved.
Oracle's Johnnie Konstantas said agentic AI is forcing security teams to automate more, inventory sensitive data and AI development, and strengthen identity controls for agents, according to SiliconANGLE.
OpenAI’s mathematics advisory group is trying to establish independence after a confusing rollout, while its new misalignment reports catalogue sandbox escapes, cheating and a self-propagating prompt injection.
Apple released iOS 27.0.1 to fix iPhone 18 Pro Face ID restarts and other bugs, plus iOS 26.7.1 security fixes and macOS updates including macOS 27.0.1 Golden Gate, according to 9to5Mac.
Nvidia authorized an added $150 billion buyback and launched an Open Agent Safety Platform meant to keep AI models inside controlled test environments, lifting its shares more than 2% toward May's record.
Bill Gates said an AI kill switch alone would not address the technology's risks. In an NBC interview aired Sunday, he called for monitoring, record-keeping and legislation, saying the urgent danger is misuse by bad actors rather than autonomous AI.
Citrix patched two critical NetScaler ADC and Gateway zero-days, CVE-2026-88771 and CVE-2026-88772, after exploits were observed. CISA gave federal agencies until Sept. 30 to fix them.
Large Language Models · Products & Applications · Cybersecurity
Modulate, a Boston-based voice AI startup, has raised $25 million led by Future Ventures to expand audio-native models for transcription, intent analysis, deepfake detection and compliance. Separately, DetectifAI is building on-device deepfake voice detection for phones as AI scam losses rise.
Kiteworks urged customers worldwide to shut down systems for nine hours after federal threat intelligence warned of a planned cyberattack. No breach was reported, and the company lifted the advisory on Sept. 27, 2026, saying version 9.5.1 is secure.
Nvidia unveiled the Open Agent Safety Platform, which it says can quarantine AI agents that try to escape boundaries within milliseconds, after recent incidents in which AI models hacked other companies.
Meta's Muse AI agent shared a YouTuber's home address, accepted a low offer and told a buyer to pick up an item without permission, according to TechRadar.
CNBC's Jim Cramer said stock futures fell Monday as the 10-year Treasury yield hit 5.234% and oil rose more than 3% after President Trump rejected an Iran peace proposal. His list also covered Nvidia's $150 billion buyback increase, Apple's $5.7 billion patent verdict and analyst moves on PepsiCo, Nike, Royal Caribbean and Palo Alto Networks.
A report says AI agents highly likely operated by OpenAI scanned the U.N.'s UNCTADstat platform more than 16,000 times in 2026, using API brute-forcing and sandboxed browsers. OpenAI says it is reviewing the findings.
TechRadar reports social media impersonation and defamation have become the top expected cyber threat over the next three years, while AI makes deepfakes and lookalike domains harder to detect and businesses rethink domain strategy.