Skip to content

Topic

Sandboxing

A security technique that restricts a program's access to files, networks, credentials, and other resources to less than its user could reach.

Current stories

productOne report1 publisher

nFOX Lab's Blockhouse sandbox starts self-hosted AI models with no privileges

nFOX Lab plans to release Blockhouse, a kernel-level sandbox that removes a self-hosted AI model's privileges by default, under the Apache 2.0 license. For now it is available as a beta for qualified enterprise evaluations.

Publishers:siliconangle.com

Reality

Evidence35
Adoption10
Hype gap+10
Incentives55
Confidence50
buildOne report1 publisher

Of 12,915 MCP servers HISTOR started in a sandbox, 41% never answered

HISTOR started 12,915 npm and PyPI MCP servers in a sandbox, and 5,259 (41%) never answered, according to its own post. About 24% were broken or not stdio servers at all. Another 16% depended on keys, flags or runtimes a client config can supply.

Publishers:dev.to

Reality

Evidence66
Adoption
Not scored
Hype gap−10
Incentives
Not scored
Confidence64
buildOne report1 publisher

AWS's Strands Box routes agent actions through a policy engine that remembers them

AWS's Strands Box preview, released October 7-8, 2026, routes every agent action through a policy engine with shared history, a dev.to post reports. With that history, teams can write rate limits and cross-tool rules that a sandbox judging one call at a time cannot hold.

Publishers:dev.to

Reality

Evidence30
Adoption
Not scored
Hype gap+15
Incentives
Not scored
Confidence30
securityConfirmed9 publishers

Nvidia adds BlueField-4 hardware watchdog to back up host-based AI agent containment

Nvidia on Monday launched an AI agent safety platform whose Sentry watchdog runs on its BlueField-4 data processors. The design assumes agents will work around their limits, so its strongest enforcement runs on hardware only Nvidia makes.

Perspective Coverage

9 publishers
Builder
Builder 35%
Operator
Operator 44%
Investor
Investor 21%

Reality

Evidence55
Adoption35
Hype gap+25
Incentives75
Confidence60
buildConfirmed2 publishers

Meta's Muse agent put its own 6.8GB Linux image, SSH key files included, in a user's Google Drive

Meta's Muse agent wrote 6.8GB of its own Linux environment, SSH key files included, to Mouse founder Peter James's Google Drive, he says. James has not tested the keys, yet the export shows anything baked into an agent's image can leave through a storage connector the user linked.

Publishers:t.signalplus.com tokenpost.com

Reality

Evidence35
Adoption
Not scored
Hype gap+25
Incentives
Not scored
Confidence40
buildOne report1 publisher

An unprivileged process can confine itself to a directory list without root

Landlock has been in mainline since kernel 5.13 and needs no policy file and no administrator to confine a process to named paths. The dev.to walkthrough explaining it leaves the enforcing syscall inside a comment.

Publishers:dev.to

Reality

Evidence30
Adoption35
Hype gap+45
Incentives30
Confidence45
buildOne report1 publisher

Chrome adds V8's in-process sandbox to its bug bounty before calling it a strong boundary

V8 says 60% of the Chrome exploits caught in the wild between 2021 and 2023 started in its own code, and that the logic bugs behind them are out of reach of both Rust and memory tagging, so the work went into containment.

Publishers:v8.dev

Reality

Evidence62
Adoption35
Hype gap−10
Incentives70
Confidence58
productOne report1 publisher

GitHub moves Copilot's sandbox lock into the JetBrains plugin itself

Admins can now set filesystem, network, proxy, developer-tool and Keychain limits for Copilot sessions from inside the plugin rather than through a device-management request, and a new diagnostics tool tells them whether the policy reached the laptop.

Publishers:devops.com

Reality

Evidence42
Adoption
Not scored
Hype gap+12
Incentives58
Confidence48