Amazon launches private AI bug bounty program for Nova

Proud to share that we’ve launched Amazon’s private AI bug bounty program — designed to strengthen the safety and resilience of Amazon Nova models. 👉 https://lnkd.in/e8JMV9dd I’m excited to see this collaboration with researchers & security experts and see this as a major step toward responsible AI — collaborating with the broader community. #ResponsibleAI #GenerativeAI #AISafety #AmazonNova

Great partnership between Amazon Nova AI Challenge (https://lnkd.in/eUbs9PBq) and Amazon Bug Bounty (https://lnkd.in/ev7CethF). Connecting leading academic teams in AI/LLM security with the professional security research community to work together to address new and emerging challenges.

Amazon has announced a new private bug bounty program focused on its Nova foundation models and AI applications. The initiative complements Amazon’s existing public AI bug bounty, which has already led to more than 30 validated security findings and over $55,000 in rewards. The new private program invites selected professional security researchers and academic teams to identify vulnerabilities that could impact the safety and integrity of Amazon’s AI systems. Areas of focus include prompt injection and jailbreak attacks, model vulnerabilities with real-world exploitation potential, and ways AI models might inadvertently enable harmful behaviour, including in chemical, biological, radiological, or nuclear contexts. The program officially began with a live launch event at Amazon’s Austin office and will expand to include more invited participants in early 2026. Rewards range from $200 to $25,000, depending on the severity and novelty of the discovered issues. By opening this private AI bug bounty, Amazon aims to reinforce transparency, collaboration, and safety in its AI systems. The company highlights that securing generative AI requires joint efforts from researchers, academics, and the wider security community — ensuring that Nova models remain robust across Amazon’s ecosystem, including Amazon Bedrock and other AI-powered products.

OpenAI quietly released Aardvark. An AI security researcher that scans repos continuously, confirms exploitability in a sandbox, drafts patches and opens PRs. Powered by GPT-5. Works with GitHub. Private beta. In testing it caught 92% of vulns, including tricky edge cases. Security work is shifting. https://lnkd.in/eSzXD7xh #infosec #security #coding

4 ways Google uses AI for security, catalog of AWS threat actor techniques, training a custom small language model to find secrets

**ML & Data-Driven Forensic Automation** 🔍🤖 I'm excited to share my latest project that combines Machine Learning with Digital Forensics! This comprehensive toolkit automates cyber investigation processes through AI-powered analysis. **Key Features:** - 🧠 **Network Traffic Analysis**: ML-based classification of benign vs malicious network patterns - 💾 **Memory Forensics**: Integration with Volatility 3 for advanced memory dump analysis - 📊 **CASE-Compliant Data Handling**: Standards-based forensic data management and interoperability - ☁️ **Cloud Ready**: Supports Google Colab for scalable forensic analysis **Tech Stack:** Python, Scikit-learn, Scapy, Pandas, NumPy This project addresses the growing need for automated anomaly detection in digital forensics, enabling investigators to process large volumes of evidence data efficiently. Perfect for cybersecurity researchers and forensic professionals looking to leverage AI in their workflows. **Repository:** https://lnkd.in/gpXeu_sx #MachineLearning #DigitalForensics #CyberSecurity #AI #Python #Volatility3 #NetworkSecurity #DataScience #InfoSec #ForensicAnalysis

Amazon has launched an invite-only bug bounty program for its NOVA family of language models, allowing select researchers to test and be paid for findings on issues such as prompt injection, jailbreaking and other vulnerabilities, with the company saying the effort will help secure models integrated across Amazon and customer systems. https://lnkd.in/eCmw7JYu #cloud #news #research #vendors #vulnerabilities #aisecurity #aivulnerability #amazon #amazonaws #amazonbedrock #bugbounty #ictinnovations #jailbreaking #promptinjection #securityresearchers

Last week, OpenAI announced Aardvark, an agentic security researcher powered by GPT‑5. It's an autonomous agent in private beta, designed to “help developers and security teams discover and fix security vulnerabilities at scale” by continuously analyzing source code. I'm super excited about this, but I found the early announcement a little underwhelming – no stats yet on efficacy other than on public benchmarks (with 92% detection, but vulnerable to overfitting), and 10 CVEs assigned from OSS projects (compared with Google's Big Sleep, which reported 20 CVEs in August). It’s a big milestone to find these autonomously, but without knowing the cost and false positive rate, it's hard to say whether Aardvark is a leap over what we've seen to date. (I appreciated that the DARPA AIxCC competition required competitors to publish cost per vulnerability). Still, the base model capabilities only get better! Looking forward to more benchmarks in this space from Aardvark, Google's CodeMender, and Anthropic's Claude Security Review.