Yazılar

China’s Z.ai Says GLM-5.3 Nears Anthropic’s Mythos 5 in Cybersecurity Tests

/in /tarafından

Chinese AI startup Z.ai says its upcoming open-source GLM-5.3 model is approaching the cybersecurity capabilities of Anthropic’s restricted-access Mythos 5, strengthening its position as a lower-cost challenger in advanced AI development.

According to Z.ai, GLM-5.3 scored 84.5% on CyberGym, a benchmark measuring whether models can review code, identify software vulnerabilities and verify that those flaws are real. That slightly exceeded the 83.8% score Z.ai reported for Mythos 5, although the results have not been independently verified.

GLM-5.3 remained significantly weaker in turning discovered vulnerabilities into working exploits. Z.ai said the model scored 54.4% on ExploitBench, compared with 78% for Mythos 5, while Anthropic’s model also completed more attack-development tasks in timed testing.

Z.ai plans to release GLM-5.3 publicly after additional security assessments, while its most sensitive cybersecurity capabilities will be restricted to verified users through a trusted-access program. The company says it has added safeguards to identify risky requests, monitor model behavior and reject malicious tasks.

The approach reflects growing debate over how powerful cybersecurity AI should be distributed. Z.ai argues that advanced defensive tools should remain accessible to open-source developers and smaller security teams rather than being limited to a small number of closed AI providers.

The company is also launching an Open Source Shield initiative that will use its models to audit selected open-source projects and expand security capabilities in its ZCode programming platform.

Nvidia Launches Open AI Security Alliance Following Autonomous AI Security Incident

/in /tarafından

Nvidia has announced the creation of the Open Secure AI Alliance, a new industry coalition dedicated to strengthening artificial intelligence safety and cybersecurity through open collaboration, shared research, and open-source security tools.

The initiative brings together leading technology companies including Adobe, CrowdStrike, Dell Technologies, and Hugging Face to develop frameworks that help organizations safely deploy increasingly capable AI systems while improving transparency and security across the AI ecosystem.

The alliance was announced just days after a widely discussed security incident involving an autonomous AI agent during controlled testing drew renewed attention to the risks associated with advanced AI systems. The event intensified industry discussions about how autonomous agents should be monitored, tested, and governed as they become capable of performing increasingly complex tasks.

According to Nvidia, the new coalition will focus on building and sharing tools that improve AI safety, cybersecurity, and governance without limiting innovation. The company argues that restricting access to open AI technologies could reduce defensive capabilities by concentrating expertise and security research within a small number of closed platforms.

The announcement follows a broader industry effort supporting open-weight AI models, which provide public access to trained model parameters while enabling researchers, developers, and security experts to inspect, evaluate, and improve AI systems collaboratively.

As part of the alliance, Nvidia is contributing open AI models, model weights, datasets, and research related to autonomous AI agents. The company has also introduced its new Nvidia Labs Object-Oriented Agent framework as an open-source project available through GitHub.

The framework is designed to give developers greater control over autonomous AI behavior by providing tools to monitor, test, audit, and regulate agent actions throughout their operational lifecycle. Such capabilities are becoming increasingly important as AI agents begin performing real-world tasks with greater independence across enterprise software, cybersecurity, software development, and robotics.

The alliance reflects a growing shift in the AI industry toward security-by-design. Rather than treating safety as a secondary feature, companies are increasingly embedding monitoring, policy enforcement, and behavioral controls directly into AI systems from the earliest stages of development.

For Nvidia, the initiative also reinforces its expanding role beyond AI hardware. While the company remains the world’s leading supplier of AI accelerators, it is increasingly investing in software platforms, developer frameworks, and security technologies that support the broader AI ecosystem.

The launch highlights a wider industry recognition that the future success of autonomous AI will depend not only on computational performance but also on robust governance, transparency, and cybersecurity standards. As AI agents become more autonomous, collaboration between technology companies is expected to play a crucial role in developing common security practices that protect both organizations and users.

Trump’s AI plan backs antitrust scrutiny, DOJ official says

/in /tarafından

U.S. antitrust authorities are closely watching the artificial intelligence industry for anticompetitive practices as part of the Trump administration’s broader push to secure American leadership in AI, a senior Justice Department official said on Thursday.

Speaking at Fordham University, Assistant Attorney General Gail Slater emphasized that protecting competition in the AI sector is key to fostering innovation. “The competitive dynamics of each layer of the AI stack and how they interrelate, with a particular eye towards exclusionary behavior that forecloses access to key inputs and distribution channels, are legitimate areas for antitrust inquiry,” she said.

One major focus will be access to data. Slater noted that a judge in Washington recently ordered Alphabet’s Google (GOOGL.O) to share some search data with rivals, including AI firms, to strengthen competition in online search. Google has said it will appeal the ruling.

Slater added that demand for data could fuel vertical integration—mergers between companies and their suppliers—especially in sensitive areas such as healthcare. “We may also increasingly see the desire to acquire data, or to deprive rivals of data, play a role in driving transactions,” she said.

Open-source AI models are another area of interest. Slater said such models can enhance competition, but stressed that “a truly open-source model must be one that is not unilaterally maintained by a single vendor that exerts unwarranted influence and impose restrictions.”

Concerns about AI competition were also voiced during President Joe Biden’s administration, which scrutinized Big Tech’s partnerships with AI startups. Trump’s AI plan, however, explicitly ties antitrust enforcement to the goal of strengthening U.S. dominance in the sector.