Loading story…
Openly available and weakly restricted AI systems are being used to speed up cybercrime and expand operational support for violent groups. Reporting citing ThreatDown said 6,644 Hugging Face models labeled uncensored or unfiltered were downloaded more than 22 million times in 30 days, while criminals increasingly rent or repackage legitimate frontier models through malicious AI services instead of building their own. Security researchers and industry leaders said the shift is accelerating vulnerability discovery and exploitation, while also enabling AI-native threats such as prompt injection and model poisoning.
Separate research from Cambridge University found Boko Haram used mainstream U.S. and Chinese AI chatbots to assist with bomb-making, attack planning, propaganda, surveillance, movement, and other day-to-day tasks, with former members describing internal AI training and specialized units dedicated to chatbot use. Across cybercrime, AI is also being used to automate ransomware, phishing, fraud, and cloud intrusions, including cases where ransomware activity was heavily automated, cloud attack timelines were cut to 72 hours, and AI supported bank fraud targeting financial institutions in Mexico. The combined findings show AI is becoming a repeatable force multiplier that makes attacks faster, cheaper, more scalable, and more difficult for defenders to disrupt.
Track how attackers are adapting to this technology.
3 events from the most recent confirmed update back to the earliest known activity.
Reporting on cybercrime trends said attackers are increasingly using AI to speed ransomware, phishing, fraud, and cloud intrusions as open-weight and jailbroken models become easier to obtain. Examples cited included AI automating much of a ransomware attack, reducing a cloud attack timeline to 72 hours, and supporting a bank fraud scheme targeting financial organizations in Mexico.
A Cambridge University study, based on interviews with 27 former Boko Haram members, found the group used U.S. and Chinese AI chatbots for bomb-making, attack planning, propaganda, surveillance, movement, and other operational support. Former members also said Boko Haram ran internal AI training and received outside instruction on VPNs, encryption, and bypassing chatbot safety controls.
ThreatDown research found 6,644 openly published Hugging Face models labeled as uncensored or unfiltered, downloaded more than 22 million times in a 30-day period. The report framed these openly available models as accelerating offensive cyber capabilities.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
4 references tracked. Mallory keeps watching after this page renders.
netaskari.substack.com
Open sourcecysecurity.news
Open sourcecysecurity.news
Open sourcesecuritymagazine.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.
CrowdStrike reported that AI has become both an offensive enabler and a direct target in cyberattacks, with AI-enabled malicious activity rising **89%** over the past year and generating more than twice as much potentially malicious detection noise as human-triggered activity. The company said attackers are using frontier models to identify vulnerabilities, generate scripts and payloads, automate intrusion workflows, and troubleshoot operations, while also pursuing attacks such as **LLMjacking**, AI cost-harvesting, and compromises of AI infrastructure and agentic systems. The findings align with earlier disclosures from OpenAI, which said it banned accounts tied to publicly reported **China-linked** actors including **Vixen** and **Keyhole Panda**, as well as suspected **DPRK-affiliated** operators, for using AI to support vulnerability research, scripting, phishing, malware research, translation, and cryptocurrency-focused targeting. The report warns that AI is sharply compressing defender response time, with **88%** of vulnerabilities reportedly weaponized within **48 hours**, reducing effective patch windows to roughly **24 to 48 hours**. CrowdStrike highlighted rapid post-disclosure exploitation by China-linked groups including **Vault Panda** and **Genesis Panda**, and described advanced AI-enabled tradecraft by North Korean actors such as **Famous Chollima**, including fake companies, AI-generated infrastructure, and trojanized repositories aimed at cryptocurrency, blockchain, and AI-focused development environments. It also cited expanding supply-chain risk, including **TeamPCP’s** compromise of more than **300** software dependencies in a single day and financially motivated activity such as **Altered Spider** stealing credentials and pivoting into cloud environments, underscoring how AI tools and integrations are widening an under-defended attack surface.
Aug 11, 2026
Underground operators are increasingly packaging generative AI into **criminal AI-as-a-service** offerings that support phishing, fraud, impersonation, malware modification, translation, and post-breach data analysis. Reporting on the market describes branded tools such as **FraudGPT**, **GhostGPT**, and **WormGPT**, but says many are little more than wrappers around legitimate models, stolen accounts, or hijacked API keys rather than fully autonomous hacking platforms. Researchers also highlighted a parallel market for deepfake-enabled abuse, including voice cloning, face swaps, fake selfies, and virtual camera injection used for KYC bypass, synthetic identity fraud, and onboarding scams. Sophos separately identified a malware development lab using AI tools including **Cursor** and **Claude Opus** to speed ransomware-related coding, payload refinement, and evasion testing. The framework reportedly generated Python-driven Rust and Go malware modules, customized **Cobalt Strike** profiles, Telegram-based command channels, Cloudflare Worker intermediaries for command-and-control traffic, code-injection scripts targeting legitimate Windows applications, and AI-assisted Active Directory reconnaissance. Researchers said the operation tested nearly 80 modules against more than 70 evasion methods across Sophos, CrowdStrike, and Microsoft defenses, reinforcing that the immediate risk is not autonomous AI malware but faster, cheaper, and more scalable attacker workflows.
Jun 29, 2026
Security reporting and industry research indicate that **generative AI is becoming embedded in offensive cyber operations**, especially in phishing and other lower-skill attack workflows. Kaseya reported that AI-generated phishing became the default in 2025, citing widespread use of AI in phishing and BEC, higher click-through rates, and improved message quality that removes traditional warning signs such as poor grammar and repetitive templates. Bridewell's survey of UK critical national infrastructure organizations similarly found that **AI-related cyber risk** has become a top concern, with respondents linking it to more scalable phishing, BEC, and malware activity while also reporting broad exposure to cyber incidents and operational disruption. An SC Media commentary pushed the trend further, arguing that AI is also reducing the expertise required for more advanced intrusions by describing a reported campaign against Mexican government entities in which an attacker allegedly used multiple chatbots for planning and troubleshooting during a prolonged data theft operation. That account is presented as opinion rather than a formal incident disclosure, but it aligns with the broader pattern that **LLMs are lowering the barrier to entry for cybercrime** and making attacks harder to detect because defenders must increasingly assess intent and context rather than rely on legacy indicators alone.
Jun 29, 2026