Loading story…
Researchers at North Carolina State University have developed RisingAttacK, a novel adversarial technique that can manipulate what AI vision systems perceive without changing the image to the human eye. This attack allows adversaries to control object identification in advanced computer vision systems. The technique highlights significant risks for AI-driven technologies in sectors like autonomous vehicles and medical imaging.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
3 events from the most recent confirmed update back to the earliest known activity.
A separate Help Net Security report emphasizes that AI development needs ethical safeguards to prevent real-world harm, reflecting broader public discussion of AI safety and governance.
Help Net Security reports on Cybersecurity AI (CAI), an open-source framework focused on AI security, indicating a new defensive development in the AI security ecosystem.
A report describes RisingAttacK as a new adversarial threat capable of manipulating what AI vision systems perceive, marking the public emergence of this specific AI security issue.
Map 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.
Multiple academic research reports describe new ways attackers can subvert or extract sensitive information from **AI systems** beyond traditional prompt-based abuse. University of Florida researchers presented *Nullspace Steering* (“**Jailbreaking the Matrix**”) as an internal, model-level technique to bypass AI guardrails by manipulating a model’s internal decision pathways, arguing that evaluating and hardening safety controls requires “popping the hood” rather than only testing external prompts. Separately, researchers from Shanghai Qi Zhi Institute, East China Normal University, Tsinghua University, and the Chinese Academy of Sciences reported that **LLM model-editing** workflows (e.g., “locate-then-edit” approaches intended to remove or correct memorized sensitive data) can leak the very data they aim to eliminate because **parameter updates act as a side channel**; they describe a two-stage reverse-engineering attack framework (*KSTER*) and a mitigation approach (*subspace camouflage*). Georgia Tech researchers disclosed a backdoor-style vulnerability dubbed **VillainNet** affecting **AI “super networks”** used in autonomous driving, where a malicious modification can remain dormant across many configurations and then trigger under specific real-world conditions (e.g., a particular subnetwork selected during rain/road changes), enabling targeted control of vehicle behavior with high likelihood of success. In contrast, a Gopher Security blog post argues for **lattice-based, zero-trust identity verification** for agentic AI integrations (including *Model Context Protocol* tooling) and warns about quantum risks to RSA/ECC and token signatures; it is largely forward-looking guidance/advocacy rather than reporting a specific newly disclosed vulnerability or incident tied to the other research items.
Jun 29, 2026
The rapid advancement of artificial intelligence is fundamentally transforming both the threat landscape and defensive strategies in cybersecurity. Attackers are leveraging AI to create sophisticated deepfakes, automate penetration testing, and develop new forms of malware that can bypass traditional security controls. Notably, a real-world incident involving the engineering firm Arup saw deepfake impersonation used to steal $25 million, highlighting the tangible risks posed by AI-powered social engineering. Security professionals are responding by developing autonomous threat-hunting tools and digital twins to counteract adversarial AI bots, but the arms race is escalating, with attackers often gaining the upper hand due to the speed and scale enabled by AI. Researchers and practitioners emphasize the need for smarter, AI-aware authentication and proactive defense mechanisms to keep pace with evolving threats. At a strategic level, experts warn that the accelerating pace of AI innovation is outstripping the ability of national security and defense systems to adapt, potentially leading to strategic surprises and undermining long-term planning. AI's ability to rapidly test and deploy new attack techniques, such as autonomous penetration testing bots that have discovered critical vulnerabilities in widely used products, is shifting the economics and dynamics of cybersecurity. Organizations are urged to rethink their security postures, invest in continuous threat hunting, and prepare for a future where AI-driven attacks and defenses operate at a velocity and complexity beyond human tracking. The consensus is clear: the AI arms race in cybersecurity is intensifying, and both attackers and defenders must evolve rapidly to survive.
Jun 29, 2026
The rapid integration of artificial intelligence (AI) technologies into business operations has introduced new risks and challenges for cybersecurity teams. High-profile incidents, such as the accidental leakage of confidential information at Samsung via ChatGPT and the manipulation of AI chatbots at companies like Air Canada and Chevrolet dealerships, have highlighted the potential for both accidental and malicious exploitation of generative AI tools. As organizations increasingly rely on AI for efficiency and innovation, experts emphasize the need to adapt incident response plans to address AI-specific risks, referencing resources like MIT's AI Risk Repository to categorize and understand these emerging threats. While AI is automating repetitive security tasks and augmenting defensive capabilities, it is also being leveraged by attackers to enhance existing tradecraft, such as phishing and code generation. Recent research and industry analysis indicate that most so-called "AI malware" remains at an early stage of maturity, serving as a force multiplier rather than introducing fundamentally new attack techniques. Security professionals are advised to focus on monitoring the abuse of legitimate AI services and to avoid overreacting to sensationalized claims of fully autonomous AI-driven attacks, as the real-world impact of such threats remains limited at present.
Jun 29, 2026