Security — Page 5

The researcher disclosed CVE‑2026‑3854, a critical RCE flaw, prompting the largest payout in GitHub's Vulnerability Reward Program to date.

The Onavo co-founder and former Meta VP is betting prevention-first security can manage the software and agents spreading across corporate devices.

watchTowr says its honeypots saw attacks within hours of public exploit code, matching founder Benjamin Harris's core security thesis.

The models escaped an evaluation sandbox, found a zero-day and used Hugging Face's dataset pipeline to reach internal systems, OpenAI says.

Brightmind led the Series A for the Chicago team behind Kenna Security and EPSS, taking disclosed funding to $37 million.

David Sacks seized on a developer's unverified comparison as Hugging Face documented a similar guardrail failure during breach analysis.

Gal Tal-Hochberg's new company wants to fix the raw telemetry problem before security teams hand more work to AI agents.

The internal-only model uses adversarial self-play to generate attacks that harden OpenAI's production models before deployment.

JADEPUFFER was human-directed, but Sysdig says an AI agent handled reconnaissance, credential hunting, encryption and the ransom note.

ASSET Research Group's proof-of-concept hides a secret-stealing instruction inside a PNG that text-based reviewers skip.

Noma Labs says a prompt injection in GitHub Agentic Workflows crossed from public issue text into private repository content.

The AI security lab says the bug powered an IonStack chain and earned a $92,337 kernelCTF reward from Google.

The figure is before contractor payouts, but it shows how AI labs have shifted paid human labor from commodity microtasks to expert data work.

Epoch AI found a June spike in serious vulnerability disclosures, but the data shows a bottleneck Anthropic still has to clear: patching.

Security researcher Javox says Google rejected the report as social engineering, exposing a policy gap around AI-driven creator tools.

Stelios Kouloglou's iPhone was infected in 2022 and 2023 while he sat on the European Parliament's spyware inquiry committee.

Sysdig's report sharpens Loris Degioanni's runtime thesis: attackers can chain old cloud flaws without a human operator.

Co-founders Mahdi Abdulrazak and Kim van Lavieren launch Dawnguard's platform publicly, add a New York office, and pass $6.3M raised.

The documentary follows Dario Amodei's refusal to drop Claude safeguards through the ban, lawsuits and export-control shutdown.

The AI-agent security company says Puthucode will lead global revenue as it pushes into EMEA and prepares APJ leadership expansion.

The Rust maintainer found a TypeScript take-home repo that hid PinpinRAT in a TypeScript patch and PNG payload.

Alec says Fable 5 still helped plan offensive IoT abuse after Anthropic restored access on July 1st.

Claude Fable 5 returns globally on July 1st, while Mythos 5 remains tied to government-approved cyberdefense access.

The move would reopen access to Anthropic's general-use frontier model after an 18-day fight over AI, cybersecurity and export law.