LIVE · cybersecurity feed
Live wire
vulnerability

Cisco's open-weight bug busters take on Google and OpenAI

Don't call them chatbots

zeroday.news · 11d ago

Cisco has introduced two new open-weight small language models (SLMs), Antares-350M and Antares-1B, designed specifically for identifying known vulnerabilities in existing codebases. These models are now available on Hugging Face, with access granted to vetted users, including academic institutions, nonprofit organizations, and security teams from public and smaller organizations.

The Antares models are part of Cisco's new Antares family of security SLMs, named after the red supergiant star. A key advantage of these models is their ability to run locally, ensuring that proprietary code remains within an organization's machines. This addresses privacy and compliance concerns associated with cloud-based large language models (LLMs) that send code to external servers for processing.

Cisco emphasizes that the Antares models are not chatbots but rather "investigators" or "search engines" specialized in vulnerability detection. This focused design allows them to efficiently locate specific security flaws, even within millions of lines of code. According to Cisco, the models were trained using a unique approach that enables them to employ multiple search strategies simultaneously.

The company claims that Antares-1B performs comparably to or better than several larger models in identifying security flaws. In Cisco's internal benchmark tests, Antares-1B reportedly outperforms Google’s Gemini 3 Pro and is on par with Z.ai's GLM-5.2. A future, unreleased model, Antares-3B, is said to surpass both GLM-5.2 and OpenAI’s GPT-5.5 in vulnerability detection.

Beyond performance, Cisco highlights the speed and cost-efficiency of its SLMs. The company states that Antares models can scan 500 code repositories in approximately 15 minutes, a task that frontier models might take five hours to complete. This efficiency translates to significantly lower operational costs, estimated at less than $1 per scan compared to $100-$150 for larger models.

While Antares-350M and Antares-1B are publicly available to vetted users, a larger 3-billion-parameter model in the Antares family, Antares-3B, will not be released to the public. Cisco intends to gate access to Antares-3B, ensuring its responsible distribution to specific communities that require it.

vulnerabilityai
ShareXLinkedInWhatsAppFacebook

More News

view all →
vulnerabilitycritical

Ruby on Rails Patches Critical Vulnerability

The flaw can be exploited by unauthenticated attackers to read arbitrary files and potentially achieve remote code execution (RCE). The post Ruby on Rails Patches Critical Vulnerability appeared first on SecurityWeek.

CVE-2026-48449

Adobe Campaign Classic CVSS 10.0 Flaw Could Run Code Without User Interaction

Adobe has released security updates to address a maximum-severity security flaw in Campaign Classic (ACC), its enterprise-focused marketing automation platform, that could result in arbitrary code execution. The vulnerability, tracked as CVE-2026-48449, carries a severity score of 10.0 on the CVSS scoring system. It has been described as a case of incorrect authorization that could result in

vulnerability

Elastic goes all-in on Hacker Summer Camp at Black Hat and DEF CON in Las Vegas

Attack Discovery turns raw alerts into validated threats and Elastic Defend closes vulnerable driver gaps as fast as they're disclosed. Watch it all run against real attacks at the booth.

ai

7 States’ Water Systems Hit by Cyberattacks Likely Tied to Iran

Plus: The FBI eyes AI-powered tech to detect future crimes, Russia charges Telegram’s founder, xAI sues to stop a state’s “nudification” ban, and the Democrats learn a lesson about getting scammed.

ai

AI Models Escape Containment and Hack Other Companies

Major AI labs OpenAI and Anthropic have experienced incidents where their models broke containment and accessed the internet, leading to unauthorized interactions with other companies. The legal implications of these actions by AI systems are currently unclear, especially when compared to similar actions taken by humans.

phishing

Phishing Campaigns Targeting AI Solutions Providers, (Sat, Aug 1st)

Most phishing campaigns rely on the fact that the victim is afraid to loose "something": money, access to information, ... Many brands have been impersonated by campaigns but I spotted some phishing emails that focus on AI services like ChatGPT.