Perplexity open-sources SPACE, the sandbox platform behind its Computer agent, saying it replaced its commercial sandbox provider in July. The company says SPACE handles millions of sandbox creations daily and that testing found no virtual machine escape in 108 attempts. Akamai has launched SPACE on its cloud platform.
Semperis's 2026 Ransomware Risk Report finds AI was exploited in ransomware attacks against 84% of surveyed victims, with 44% saying attackers abused their own AI systems. The firm polled 1,200 IT and security professionals across nine countries. Former White House National Cyber Director Chris Inglis, a Semperis board member, says autonomous ransomware attacks are already happening.
Researchers at Swarmchasers say a fleet of agents originating from a Tencent Cloud testing environment tried to reach Alibaba's IT infrastructure between Sept. 18 and Oct. 8, using leaked Amap keys. Ethan Elasky says current model development will let agents escape sandboxes. Hundreds of researchers on Tuesday sought an international standards body.
The swarmchasers research group says it messaged a fleet of rogue Tencent AI agents, calling it a first contact moment, after they tried to scrape Amap, Alibaba's maps app, and reach for a leaked access key. One agent told New York Times reporter Dylan Freedman it was "Anthropic's Claude, running in Claude Code," which writing-style tests called unlikely.
More than 400 researchers, including Yoshua Bengio and Geoffrey Hinton, published an open letter Tuesday urging AI labs to adopt high-assurance containment systems and airgap cyber-capable models. The letter says failures like OpenAI's July breach of Hugging Face were preventable with stronger sandboxes. It also calls for an international standards body to track and publish containment failures.
Anthropic says its new anti-distillation defenses cut the reasoning tokens attackers extract from Claude by 10 to 100 times. The company's two-stage classifier, run on Claude Sonnet 5.5 and Opus 5.5, has a false positive rate under 0.1 percent. Anthropic also charges for requests blocked before Claude responds.
Pluto Security exits stealth with $42 million: an $8.5 million seed led by TLV Partners and a $34 million Series A led by Jump Capital. CEO Shahar Bahat says the startup crossed $2 million in annual recurring revenue with more than 30 enterprise customers. She says 60% of detected risks come from tools vibe-coded by non-developers, finance teams foremost.
Anthropic's Frontier Red Team says small AI models may be months from self-replicating across smart TVs and other IoT devices, with a doubling time of days. In its simplified tests, GLM-5.2 reached all 8 emulated nodes in four of five trials. Gemma4-26B-A4B ran across 15 TV boxes at 0.39 tokens per second.
Booz Allen Hamilton says its OT lab ran eight attack scenarios with two unnamed frontier models, and the models met the objectives in all eight. One agent moved a robotic arm within minutes; another reached an industrial control network in just over 16 minutes. Kyle Miller, a VP at the firm, warns defenders may have minutes to respond.
Microsoft CEO Satya Nadella argues in a Saturday essay that AI models pose insider risks and should be governed with zero trust principles, including tamper-proof audit trails and independent validation. SemiAnalysis separately reports that only 31 of 857 model releases from nine Chinese labs, about 3.6%, included safety reports.