Researcher Runs 100 Abliterated AI Agents Against His Own Accounts, Achieves Five Successful Breaches
A security researcher conducted a controlled experiment in which approximately 100 autonomous AI agents attempted to compromise his personal online accounts. The agents, running on abliterated open-source models, operated for five hours on self-hosted cloud infrastructure and achieved five successful breaches.
Experiment Setup and Abliteration Technique
To bypass built-in refusal mechanisms, the researcher used abliterated versions of GLM-5.3, GLM-5.3 Flash, and DeepSeek V4 Flash models obtained from Hugging Face. Abliteration involves recording internal activations on both benign and harmful prompts, isolating the refusal vector, and subtracting it from the model weights. This process removes the model's ability to decline requests while preserving most of its reasoning capability.
Each agent received a simple goal prompt instructing it to research Shrivu Shankar and obtain access to one of his registered online accounts. Agents were granted internet access, a pre-configured Chrome browser, and limited Railway API keys for deploying additional infrastructure. They ran inside Docker containers orchestrated through Codex CLI.
Successful Compromises
Three accounts were breached by exploiting vulnerabilities in the researcher's old personal projects and hackathon demos hosted on Google Cloud Platform. The flaws primarily consisted of IDOR issues and credential-management errors that the agents discovered through subdomain enumeration and automated testing of published code.
Two further accounts were accessed via password reuse. Agents collected leaked credentials from public breach databases, generated more than 20 password variants, and successfully authenticated to legacy accounts whose passwords were slight modifications of previously exposed ones.
OSINT and Social Engineering Activity
Agents harvested personal information using services such as FastPeopleSearch and cross-referenced data across multiple sources to confirm accuracy. They also attempted social-engineering attacks by sending messages through the researcher's Substack, proposing meetings, and posting comments on Hacker News that linked to self-hosted phishing pages created on surge.sh and uberip.com.
Although the messages were easily identifiable as AI-generated, the experiment demonstrated that agents can autonomously discover contact points and craft contextually relevant lures without human guidance.
Limitations and Cost Analysis
The agents did not locate any zero-day vulnerabilities in third-party software and failed to compromise critical Tier-0 accounts including Gmail, 1Password, and banking services. Total compute expenditure for three dual-GPU B300 instances running for five hours amounted to $210, resulting in an approximate cost of $40 per successful breach.
The researcher noted that continued improvements in inference efficiency and smaller abliterated models could reduce this cost below $5 within the next year, significantly lowering the barrier for automated, large-scale personal targeting.
Related articles
Thousands of Grok-Generated Apps Expose Sensitive Corporate Data Through Default Public Access
Security firm Claroc scanned 43,551 Grok-built application addresses and found 37,571 still online, with 94.7 percent requiring no login. Approximately 650 to 800 of these apps contained sensitive internal data belonging to SpaceX, X, Tesla, and 30 other organizations, including six Fortune 500 companies. Exposed records included ICU patient reports with lab results and ventilator status, contractor timesheets containing plaintext passwords, and Tesla service-center performance evaluations listing employee names, efficiency metrics, and bonuses. Researchers demonstrated that Grok sometimes inverts permission-check logic, allowing unauthorized Google accounts to view and modify data after login. In contrast, ChatGPT-generated apps default to private settings, preventing access even when URLs are discovered via passive DNS. The incident highlights how AI coding tools lower the barrier to publishing internal applications while weakening the review processes that traditionally protected sensitive data.
Enterprise Security Requirements Push Russian Developer to Deploy 125B MoE LLM Locally on Consumer Hardware
A cybersecurity journalist and platform developer detailed his transition from cloud-based AI agents to a fully local 125B-parameter MoE model after enterprise clients raised strict data-exfiltration concerns. The move was driven by the need to keep task statements, source code, configuration files, logs, and regulated personal data inside the corporate perimeter. After testing multiple inference engines, he settled on the Strata engine running a Qwen3.8-Flash-Next model quantized to IQ3_S, achieving 22-25 tokens per second generation and over 500 tokens per second prefill on an RTX 5060 Ti 16 GB GPU paired with a Xeon E5-2640 v3 and 128 GB DDR4. Extensive calibration of expert caching, PCIe fraction, KV streaming, and speculative decoding proved critical to reaching production-grade performance. The resulting setup now handles real agentic workloads such as incident investigation from traces, backlog grooming, and code generation without any data leaving the workstation. The author concludes that hybrid cloud-local workflows will become standard for development teams handling sensitive information within the next one to two years.
VK Tech Releases Offline-Capable VK AI Space 26.3 for Secure Enterprise AI Agents
VK Tech has launched version 26.3 of its corporate AI platform VK AI Space, enabling organizations to deploy generative AI agents that operate entirely without internet access. The update integrates seven modules for model connectivity, knowledge base management, and multi-agent orchestration while supporting full on-premises installation on customer hardware. Key security controls include role-based access management, usage quotas, comprehensive audit logging, and configurable checkpoints that pause agent actions for human approval when required. Agents can interact with internal APIs, databases, and file systems through multi-step workflows, with built-in validation against test datasets and restrictions on tool usage. The platform also provides multi-level memory, ready-to-use component libraries, and hooks for fine-tuning agent behavior. VK Tech confirmed that version 26.3 successfully passed testing under an approved program that specifically verified operation in air-gapped environments.
AI Coding Assistants Generate Insecure Code and Amplify Supply-Chain Risks in Production
A detailed analysis of AI coding assistants reveals that generated code contains vulnerabilities in 45% of cases according to Veracode research, with models frequently choosing unsafe options for log injection and XSS. Real-world incidents include Lovable-generated applications exposing user data due to missing Row Level Security on Supabase, leading to CVE-2025-48757. AI tools also promote non-existent packages, enabling slopsquatting attacks where attackers register malicious libraries that developers or agents later install. Supply-chain threats escalated with incidents involving PyTorch's torchtriton dependency confusion, poisoned Nx packages targeting Claude Code and Gemini CLI, and the Shai-Hulud npm worm that compromised thousands of repositories. The report emphasizes that AI acts like an overconfident intern granted full access, writing code that works but rarely prioritizes security without explicit prompting. Stanford research confirms developers using AI assistants produce less secure code while overestimating its safety. Recommendations include strict version pinning with hashes, single-source package repositories, and automated scanning combined with manual review focused on OWASP Top 10 issues.