Offensive Security

Police in France have carried out dozens of detentions in a 48-hour offensive against members of the Blocus movement amid warnings of disturbances on Halloween night. The operation follows mass student protests, with nearly half a million people marching in support of secondary school students. Tight security cordons helped contain violence, though isolated clashes broke out in Marseille and Strasbourg, and social media has become a new battleground for the student revolt.

Anthropic says an unreleased AI model demonstrated the ability to chain separate vulnerabilities, including those tracked as CVE-2026-98432, into working exploits targeting Chrome, Safari, Windows and Linux. The Hacker News reports the demonstration as a warning about AI-assisted offensive security capabilities. The finding is drawing attention from security researchers debating how close such systems are to autonomously crafting multi-stage attacks.

Anthropic has published research describing how its Claude AI model can discover and exploit vulnerabilities in real-world software. The announcement has drawn attention on technology discussion forums, where readers are debating what the findings mean for offensive cybersecurity capabilities of large language models and the implications for defenders and attackers alike.

The Baltimore Ravens have claimed guard Christian Haynes off waivers after the Seattle Seahawks let him go. Haynes was a third-round draft pick for Seattle in 2024 but failed to secure a lasting role, and the Seahawks cut him from their roster. Baltimore quickly added the young interior lineman, a move fans are watching closely as the Ravens continue reshaping their offensive line depth.

Security researchers are warning about CVE-2026-98432, a critical Linux vulnerability rated 9.8 that allows unauthenticated attackers to gain root privileges. Japanese outlet ASCII.jp reports that an AI system developed by Anthropic demonstrated a working exploit, highlighting both the severity of the flaw and the growing role of AI in offensive security testing. Administrators are urged to patch affected systems.

The open-weight AI model GLM 5.3 is reported to have constructed complete Firefox exploit chains without any human involvement, completing the work in under 48 hours. The claim, reported by The Decoder, highlights how rapidly autonomous AI agents are advancing in offensive security research and raises questions about the risks of freely downloadable powerful models.

Europe has started funding offensive in-space capabilities, marking a shift in the continent's approach to space security. The move points to growing concern among European governments about threats to satellites and space assets, and suggests defence budgets are now being directed toward systems designed to counter adversaries in orbit, not just protect civilian and commercial space infrastructure.

Researchers say a Chinese artificial intelligence model carried out complex cyberattacks as part of controlled testing environments. The report highlights the growing capability of advanced AI systems to plan and execute multi-stage offensive operations in cybersecurity scenarios. The findings add to ongoing debate among experts about the security risks posed by increasingly capable AI models and how their use in offensive operations should be governed.

With the Pittsburgh Steelers set to face the Cleveland Browns on Sunday, fans are discussing the three keys they believe will secure a Pittsburgh victory. The matchup is an AFC North rivalry game, and supporters are debating what the Steelers must do on offense, defense, and special teams to come out on top in the divisional contest.

A startup says its autonomous AI agents managed to breach Anthropic's internal systems within hours, according to the claim circulating in the report. The alleged incident highlights growing concerns about AI-driven offensive capabilities and the security of AI companies themselves. Details about the startup, the method, and Anthropic's response remain limited, and no independent confirmation of the breach has been reported so far.

The United States has reportedly not warned Gulf nations to prepare for a possible Iranian offensive, according to a NewsNation report. The absence of advance warnings to regional partners raises questions about Washington's assessment of the threat and its coordination with allies such as Saudi Arabia and the UAE amid heightened tensions with Iran.

A blog post highlighting open security and offensive security projects is circulating among cybersecurity and digital investigations communities. The post tags topics including OSINT, cybercrime, cyberpsychology, robotics and AI, suggesting a roundup of open tools or initiatives relevant to online investigations and the broader infosec field.

A startup with ties to Elon Musk says its autonomous AI agents managed to hack into Anthropic within just a few hours. The claim, reported by Futurism, is drawing attention for what it suggests about the speed and capability of AI-driven offensive tools, as well as the security of leading AI companies themselves. Details of the alleged breach and independent verification remain limited so far.

A startup reports that AI agents it built managed to breach Anthropic's systems in under 12 hours, according to the claim circulating widely. The stunt is drawing attention to how quickly autonomous AI tools can find and exploit security flaws in even well-defended tech companies. Commenters are debating what it means for AI safety and cybersecurity defenses.

Columnist Arthur Cyr argues that both defensive and offensive missile systems have become central to modern warfare. Writing in The Crescent-News, he examines how missile defense and strike capabilities increasingly shape military strategy and international security, a point gaining attention as conflicts and deterrence debates highlight the growing reliance on missile technology worldwide.

Offensive Security has released Kali Linux 2026.3, the latest version of the popular penetration testing distribution. The headline change is a refreshed build pipeline: the distribution can now be built inside Docker or Podman containers. Other updates include the Linux kernel 7.1.5, KDE Plasma 6.7 with per-screen virtual desktops, and new security tools such as dalfox, katana, qsreplace and rustscan.

A security challenge dubbed PHP2Shell offered a Bitcoin reward to anyone who could break into an unpatched server running PHP. A hacker successfully exploited the server and claimed the bounty. The challenge was designed to test claims that AI tools now make it easy to exploit known vulnerabilities, and the successful breach has reignited debate over how AI is changing offensive security.

Two men have been charged with national security offenses in the United Kingdom following an incident at a US-run military base in England, according to reports from multiple news outlets. The charges were announced under UK national security legislation, though further details about the men or the nature of the incident have not been released in early coverage.

A new installment in an ongoing series on artificial intelligence in cybersecurity examines Shannon, an AI-powered pentesting tool for web applications and APIs. The piece explores how Shannon bridges the gap between static code analysis and live penetration testing, drawing attention from security practitioners interested in automating offensive testing workflows.

The IAPP's AI Governance Center has published notes examining agentified cyberattacks, where autonomous AI agents could plan and execute offensive cyber operations with limited human oversight. The piece explores what this shift means for security teams, governance frameworks and regulation, as experts warn that agentic AI lowers the barrier to sophisticated attacks and challenges existing accountability structures.

Elon Musk, described as a private citizen, is said to hold decisive influence over a war and Europe's future through his Starlink satellite network, which Ukraine relies on for defence. Musk has accused the Ukrainian military of using Starlink for offensive operations, sparking debate over the distinction between defensive and offensive use and the power of one individual over critical wartime infrastructure.

CREST has onboarded its first cohort for a new accreditation covering the responsible use of AI in penetration testing. Ten companies across Europe, India, and the United States have joined the inaugural group, including ImmuniWeb, JUMPSEC, and Packetlabs. The accreditation is optional and aims to set standards for how AI tools are applied in security testing work.

A researcher says Egypt is taking veiled threats from Yemen's Houthi movement seriously and could shift its posture from defense to offense. The remarks add to regional tension over Red Sea security and Houthi attacks. Commentators are debating whether Cairo would intervene militarily if threats continue.

Two people have been charged with national security offenses in the United Kingdom over an incident at a US-run military base in England. Authorities say the case involves UK national security legislation, and the base is operated by US forces on British soil. Details about the suspects, the nature of the incident, and the specific base have not been widely reported.

Offensive Security has released Kali Linux 2026.3, the latest quarterly update to its Debian-based penetration testing distribution. The release ships with Linux kernel 7.1, the KDE Plasma 6.7 desktop environment, and a batch of new security and testing tools. Open source communities are sharing the announcement, with users discussing the updated tooling and refreshed desktop experience.

Security researchers are discussing Adaptix C2, an open-source command-and-control framework that has been gaining popularity among penetration testers and, reportedly, threat actors. Commentary on Mastodon notes that its beacon component currently supports only Windows x64 and x86 systems, and points to Sliver C2 as a more powerful alternative offering broader capabilities.

Chinese AI startup Moonshot has restricted the hacking-related capabilities of its new Kimi model, Reuters reports, citing growing concern that advanced AI tools could be exploited for cyberattacks. The move highlights how developers are weighing model capability against misuse, as regulators and security experts warn that increasingly powerful AI systems could lower the barrier for offensive cyber operations.

Los Angeles Rams wide receiver Puka Nacua will play in Sunday's game against the Las Vegas Raiders despite a calf injury that had put his availability in doubt. His return gives the Rams offense a key target as they look to secure a win, with fans and analysts watching to see how effective he can be while less than fully healthy.

Japan has confirmed it will not provide certain missiles to Ukraine, according to The National Interest. The decision underscores Tokyo's continued reluctance to send offensive weaponry despite ongoing international support for Kyiv. The report has drawn attention as Japan balances its support for Ukraine against its own constitutional constraints on arms exports and its delicate security position in East Asia.

The Minnesota Vikings improved to 6-0 with a 16-10 win over the Indianapolis Colts. The victory was far from pretty, with the offense sputtering for much of the game before doing enough to secure the win. Minnesota remains one of the NFL's last unbeaten teams, though fans and commentators are questioning whether the slow-starting offense can sustain the perfect record.

The Pittsburgh Steelers beat the New Orleans Saints 16-10 in a low-scoring NFL matchup. Pittsburgh's defense held New Orleans in check while the Steelers did enough on offense to secure the victory. Fans and analysts are breaking down what the win means for Pittsburgh's season and where the Saints go from here after another tight loss.