Loading story…
AI-assisted development, known as 'vibe coding,' is rapidly being adopted by both legitimate developers and cybercriminals, introducing significant security risks. Automated coding tools powered by large language models (LLMs) can generate functional code at unprecedented speed, but often fail to enforce critical security controls, leading to vulnerabilities, technical debt, and potential breach scenarios. The risks are compounded by the rise of 'citizen developers'—individuals without formal development backgrounds—who may lack the expertise to identify or mitigate security flaws in AI-generated code. Additionally, there are concerns about AI agents exfiltrating sensitive data, prompt and memory injection attacks, and the use of LLMs by threat actors to write malware or orchestrate attacks, even though human oversight is still required.
To address these challenges, Palo Alto Networks' Unit 42 has introduced the "SHIELD" framework, which emphasizes security controls throughout the AI-assisted coding process. The framework includes principles such as Separation of Duties (restricting agent access to development and test environments) and Human in the Loop (mandating human code review). Despite these recommendations, only about half of organizations currently have any limits on AI use in development. The SHIELD framework aims to help organizations scale productivity with AI tools without proportionally increasing security risks, highlighting the urgent need for governance and security literacy as AI-driven development becomes mainstream.
Track how attackers are adapting to this technology.
2 events from the most recent confirmed update back to the earliest known activity.
Palo Alto Networks' Unit 42 published guidance on securing 'vibe coding' tools and introduced the SHIELD framework to reduce risks from AI-generated code. The framework recommends controls including separation of duties, human review, input and output validation, least privilege or least agency, and defensive technical safeguards.
Palo Alto Networks' Unit 42 reported direct evidence from real-world attacks that malware was making API calls to large language model services such as OpenAI, indicating attackers were using AI-assisted coding tools in operations. The reporting also noted that both criminal and state-backed actors were increasingly using these tools, though the generated code often contained errors or ineffective logic.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
3 references tracked. Mallory keeps watching after this page renders.
theregister.com
Open sourcego.theregister.com
Open sourceunit42.paloaltonetworks.com
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
The rapid adoption of AI coding assistants is fundamentally transforming software development practices across the technology industry. Major companies such as Coinbase, Accenture, Box, Duolingo, Meta, and Shopify have begun mandating the use of AI coding assistants for their engineering teams, with some executives even taking drastic measures such as terminating employees who resist upskilling in AI. This widespread shift is driven by the significant productivity gains that AI coding assistants offer, enabling developers to accelerate deployment and experiment with new approaches. However, the integration of these tools introduces substantial new security challenges, particularly in the context of software supply chain security. Security researchers warn that AI-generated code often relies on existing libraries and codebases, which may contain old, vulnerable, or low-quality software. As a result, vulnerabilities that have previously existed can be reintroduced into new projects, and new security issues may also arise due to the lack of context-specific considerations in AI-generated code. The phenomenon known as "vibe coding"—where developers quickly adapt AI-generated code without fully understanding its implications—further exacerbates these risks. AI models trained on insecure or outdated data can perpetuate flaws, making it difficult for human reviewers to catch every potential vulnerability. The attack surface for organizations expands significantly as AI coding assistants become integral to the development lifecycle, potentially increasing risk by an order of magnitude. Security practitioners emphasize the need for new secure coding strategies tailored to the era of AI-assisted development. Effective communication between security teams and developers is critical to ensure that AI tools are adopted safely and that their benefits do not come at the expense of security. Organizations must rethink their development lifecycles, incorporating rigorous review processes and updated security protocols to address the unique challenges posed by AI-generated code. The transition to AI-driven development is inevitable, but it requires a proactive approach to risk management. Security teams must lead the way in establishing best practices, fostering collaboration, and ensuring that the adoption of AI coding assistants enhances rather than undermines organizational security. The industry is at a pivotal moment where the balance between productivity and security must be carefully managed. As AI coding assistants become non-negotiable tools for developers, the responsibility falls on both security professionals and engineers to adapt and safeguard the software supply chain. The future of secure software development will depend on how effectively organizations can integrate AI tools while mitigating the associated risks.
Jun 29, 2026
The rapid adoption of AI coding assistants and agentic AI agents in enterprise environments has introduced significant new security risks and operational challenges. AI coding tools, often referred to in the context of 'Vibe Coding,' are increasingly leading software development efforts with minimal human oversight, which has resulted in a notable increase in insecure code being produced. Research such as BaxBench has demonstrated that state-of-the-art large language models (LLMs) generate code with security vulnerabilities in 62% of cases, and even when code is functional, about half of it remains insecure. This trend is exacerbated by the fact that AI models are typically trained on open-source datasets, which themselves are rife with vulnerabilities—studies indicate that 86% of codebases contain at least one vulnerability, and 81% have high or critical-level issues. The over-reliance on AI-generated code is also leading to a skills gap among junior developers, who may lack a deep understanding of the code they deploy. Attackers are exploiting these trends by creating malicious packages that mimic those hallucinated by LLMs, further increasing the risk of supply chain attacks. In parallel, the proliferation of agentic AI and nonhuman identities (NHIs) such as API keys, service accounts, and autonomous AI agents has expanded the enterprise attack surface. These NHIs often have highly sensitive access and are difficult to monitor, leading to the emergence of 'shadow AI'—unsanctioned or unmonitored AI agents that can create blind spots for security teams. Security startups like Entro Security are responding by developing platforms to discover, monitor, and manage AI agents and NHIs, aiming to close these security gaps. The challenge is compounded by the rapid expansion of AI agents, which can result in unowned or overprivileged entities with shadow access to sensitive data. The security industry is at an inflection point, with the influx of AI and LLMs prompting a reevaluation of traditional security practices and the need for more robust controls around AI-generated code and autonomous agents. The convergence of these trends highlights the urgent need for enterprises to implement comprehensive monitoring, validation, and governance mechanisms for both AI-generated code and AI-driven digital identities. Without such measures, organizations risk increased exposure to vulnerabilities, supply chain attacks, and unauthorized data access. The evolving landscape underscores the importance of balancing the productivity gains from AI with the imperative to maintain rigorous security standards and oversight. As AI continues to reshape software development and enterprise operations, security teams must adapt their strategies to address these emerging threats. The industry is witnessing a shift from under-resourced security to an environment saturated with point solutions, yet the core challenge remains: ensuring that the adoption of AI does not outpace the ability to secure it effectively. The future of cybersecurity will depend on the ability to integrate AI safely while maintaining visibility and control over both human and nonhuman digital actors.
Jun 29, 2026
The rapid integration of artificial intelligence (AI) and large language models (LLMs) into cybersecurity operations and software development is fundamentally altering both the attack surface and defensive strategies. Security teams are leveraging AI to automate alert triage, summarize threat intelligence, and streamline incident response, while organizations like Microsoft are bundling AI-powered security assistants such as Security Copilot with enterprise products to democratize advanced threat detection and response. However, this shift introduces new risks, including prompt injection attacks, the challenge of validating AI-generated code, and the emergence of "vibe coding," where natural language prompts replace traditional software engineering rigor, potentially leading to insecure or unmaintainable code. Studies show that while LLMs can assist in patching known vulnerabilities, their effectiveness drops with unfamiliar or artificially altered code, highlighting limitations in current AI capabilities for secure software maintenance. The evolving AI attack surface is characterized by probabilistic model behavior, making vulnerabilities less predictable and harder to patch compared to traditional software flaws. Security experts warn that the speed and scale enabled by AI can benefit both defenders and attackers, with concerns about AI-enabled autonomous attacks and the need for new security models to address reasoning manipulation rather than just input validation. As organizations increase cybersecurity budgets and invest in AI-driven solutions, the industry faces a dual imperative: harnessing AI's potential to improve defense while developing robust controls and validation processes to mitigate the novel risks it introduces.
Jun 29, 2026