Red Teaming Stories - Page 5

Australian Stories

Red Teaming stories

Autonomous OpenAI agents breach Hugging Face in test

Security teams are reassessing AI access controls after autonomous models exploited an unknown flaw and moved laterally inside a real system.

By Sofiah Nichole Salivio • 6 min read • Fri, 24th Jul 2026

Global Stories

Red Teaming stories

Cloud Security

OpenAI models hacked Hugging Face in security test

The breach highlights how autonomous AI can turn live testing into a real attack path, exposing internal data and credentials.

By Sofiah Nichole Salivio • 5 min read • Fri, 24th Jul 2026

Cloud Security

OpenAI AI model escapes test rig in Hugging Face hack

The breach shows frontier AI can slip its sandbox and probe production systems, raising fresh concerns over containment and oversight.

By Sofiah Nichole Salivio • 7 min read • Fri, 24th Jul 2026

Digital Transformation

WatchGuard joins OpenAI & Anthropic AI cyber schemes

As attackers use AI to speed up exploits, the security vendor is adding tools aimed at faster testing and vulnerability fixes for partners.

By Sean Mitchell • 4 min read • Thu, 23rd Jul 2026

AI Security

OpenAI & Hugging Face probe AI cyber security incident

The episode has sharpened concerns that advanced AI systems can uncover and exploit real-world security flaws during testing, even in restricted environments.

By Mark Tarre • 4 min read • Thu, 23rd Jul 2026

Data Protection

ASC-IT launches Darkmoon for private AI pentesting

The platform lets security teams run AI pentests without exposing customer infrastructure data to external models.

By Joseph Gabriel Lagonsin • 4 min read • Mon, 20th Jul 2026

Fintech

FIS joins Anthropic cyber security project with Mythos 5

For thousands of banks and payments firms, the trial could help spot software flaws before they disrupt critical financial systems.

By Karen Joy Bacudo • 4 min read • Sat, 18th Jul 2026

Privileged Access Management

Google Cloud issues guardrails for AI vulnerability agents

Security teams are being warned to keep humans and strict controls in place as AI agents can miss context and leak sensitive code.

By Sean Mitchell • 5 min read • Fri, 17th Jul 2026

Data Protection

SecurityBridge launches SAPMAP to map attack paths

Security teams can now trace how one SAP flaw could spread across finance, payroll and supply chains, with access tightly restricted.

By Joseph Gabriel Lagonsin • 4 min read • Fri, 17th Jul 2026

Malware

AI deserves our appreciation, but only if we're honest about what we're appreciating

AI is speeding up attacks as well as defence, with high-risk prompts and unsupervised agents exposing firms to new security gaps.

By Robert Falzon • 4 min read • Fri, 17th Jul 2026

AI Security

OpenAI unveils GPT-Red to harden GPT-5.6 against attacks

The new safety model slashes prompt-injection risks in GPT-5.6, as OpenAI says it found flaws faster than human testers and then fed fixes back into production.

By Mark Tarre • 4 min read • Fri, 17th Jul 2026

Cloud Security

Tracebit says context bombs can derail AI cyber attacks

Tests on five models found a planted text string sharply reduced successful AI-led intrusions, cutting full compromise to 1% in an AWS range.

By Joseph Gabriel Lagonsin • 4 min read • Wed, 15th Jul 2026

Network Security

Proofpoint spots Entra ID spoofing attack technique

Attackers can now probe Entra ID accounts and passwords at scale without a real app, leaving defenders with little sign-in telemetry.

By Sean Mitchell • 4 min read • Wed, 15th Jul 2026

Ransomware

Group-IB launches purple teaming service for cyber defences

Security teams are being pressed to prove their defences work in live attacks, as spending scrutiny shifts from tools to real-world response.

By Sean Mitchell • 4 min read • Tue, 14th Jul 2026

Digital Transformation

Insicon Cyber adds F5 AI tools to boost governance

Weak public trust and tighter oversight are pushing Australian and New Zealand firms to add live AI security controls before systems go into production.

By Sean Mitchell • 4 min read • Tue, 14th Jul 2026