AI Safety 13

AI startup creating innovative AI solutions and services. TUNiB is a startup that creates new connections and innovations with AI technology. They build a smarter and more responsible AI ecosystem through 'DearMate,' a next-generation SNS where AI and people come together, and AI ethics guardrail solutions, essential safety devices in the generative AI era. TUNiB also offers services such as sLLM, Multi-Persona Chatbots and NLP APIs.

Teammately: AI Agent for AI Engineers, automates LLM app development. Teammately is an AI Agent for AI Engineers, designed to automate and fast-track every step of the LLM app development lifecycle, from Dev to Ops. It aims to elevate the Time-to-value, Reliability, and ROI of LLM Apps.

AI assistant for clinical documentation, revenue enhancement, and utilization management. Pieces Technologies offers an AI assistant for clinical documentation, revenue enhancement, and utilization management across inpatient and outpatient settings. Their physician-led approach develops ensemble AI solutions to assist care teams, focusing on safety and reliability. The platform integrates seamlessly into EHR systems, providing tools for on-time documentation, billing opportunity identification, and improved discharge summaries.

Adversarial verification platform for AI agents to find security and logic gaps. Fabraix is an advanced adversarial verification platform designed specifically for AI agents and multi-agent systems. Built by ex-Meta engineers, it provides a dedicated adversarial staging environment called Nyx that launches over 1,000 automated strategies to stress-test AI systems across security, logic, and alignment. Unlike traditional software testing, Fabraix uses a pure blackbox approach that adapts in real-time to the agent's behavior, helping developers find reasoning gaps, prompt injections, and reward hacking before they reach production. It supports various modalities including text, voice, and browser-based interactions, and integrates directly into CI/CD pipelines for continuous safety coverage.

macOS app factory for creating sandboxed, native terminal apps for TUI programs. Multitui is a macOS app factory designed to convert Terminal User Interface (TUI) programs into dedicated, native macOS applications. It allows users to wrap command-line tools like Claude Code, lazygit, htop, and vim into individual .app bundles with their own dock icons and window management. A standout feature is its integrated sandboxing, which leverages macOS's built-in sandbox-exec to provide fine-grained filesystem protection. This ensures that CLI tools or AI agents can run with restricted permissions, preventing them from deleting or modifying sensitive data unless explicitly allowed. Each app maintains its own shell history, snippets, and visual themes, providing a first-class macOS experience for power users.

AI safety and research company building reliable, interpretable, and steerable AI systems. Anthropic is an AI safety and research company dedicated to building reliable, interpretable, and steerable AI systems. They offer AI models like Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku, which set industry benchmarks in reasoning, math, coding, multilingual understanding, and vision. Anthropic provides AI solutions for various use cases, including collaboration, coding, and customer support, with a focus on integrity, security, and trustworthiness.

Adversa AI secures AI systems from cyber threats, privacy issues, and safety incidents. Adversa AI is a company focused on securing Machine Learning (ML), Artificial Intelligence (AI), and Large Language Models (LLM). They offer platform solutions for secure AI awareness, assessment, and assurance. Adversa AI helps organizations protect their AI systems from cyber threats, privacy issues, and safety incidents. They provide research, compliance, and industry-specific solutions to mitigate AI risks across various sectors, including finance, insurance, automotive, biometrics, and more.

Decentralized AI platform for scalable and accurate AI solutions. OpenPeer AI is a decentralized AI platform built using LonScript, the Decentralized-Internet SDK, Mojo, and the LLM Standard. It leverages BOINC as its foundation and aims to provide a scalable platform for AI, ensuring accuracy and predictability. It focuses on utility-oriented AI solutions, supporting diffusion models, generative art, AI safety, horizontal scaling, and Robotic Process Automations (RPAs).

Making reasoning AI accessible through reinforcement learning for smart and safe machines. General Reasoning is dedicated to making state-of-the-art reasoning AI more accessible. They focus on developing machines that are smart and safe through reinforcement learning, aiming to help humanity. Their work explores the frontier of machine reasoning, with a focus on scaling RL compute and open-sourcing data and models.

OpenAI's AI initiative to improve reasoning, problem-solving, and long-term task execution. Strawberry is the codename for OpenAI’s latest AI initiative focused on improving reasoning, problem-solving, and the ability to perform complex, long-term tasks, particularly in areas where current models like GPT-4 struggle. It is designed to enhance AI’s ability to handle complex mathematical problems, programming tasks, and deep research, including tasks that require long-term planning and action. OpenAI has demonstrated Strawberry’s capabilities to U.S. national security officials, indicating its potential impact and the importance placed on ensuring its safety and reliability. One of the core innovations of Strawberry is its ability to improve the reasoning capabilities of AI. This could allow the model to perform tasks that require human-like reasoning, such as solving intricate word puzzles or making strategic decisions. This project also ties into OpenAI’s broader goals of advancing AI safety and reducing errors like hallucinations in AI responses by generating high-quality synthetic data for training future models. Strawberry could play a crucial role in the development of OpenAI’s next major model, codenamed “Orion,” which is expected to be a significant advancement over the GPT-4.

Autoblocks: A GenAI platform for building, testing, and deploying reliable AI applications. Autoblocks is a GenAI product platform designed to help teams build, test, and deploy reliable AI applications. It offers a collaborative cloud-based workspace for understanding user interactions, iterating on prompts, and shipping code changes with confidence. Autoblocks provides tools for seamless collaboration, accurate evaluations, and streamlined workflows, enabling teams to deliver AI solutions with high standards of quality and safety.

Wisent provides unprecedented control over AI models through representation engineering. Wisent uses representation engineering to give you complete and simple way of steering and controlling your AI models. Whether you want to make AI safe, hallucination-free, happy or depressed, Wisent has you covered. Wisent uses representation engineering to give you unprecedented control over AI capabilities. At Wisent, we're pioneering representation engineering to give you unprecedented control over AI capabilities. Representation engineering changes that by giving you visibility into the internal workings of AI. By understanding how AI processes information, we can precisely modify its behavior, enhance its capabilities, and create AI systems that better serve your specific needs. With Wisent's technology, you can transform AI from a rigid tool into a flexible, adaptable system that evolves with your needs. Representation engineering gives you unprecedented control over AI behavior with precision and ease.

SWMS AI automates safety compliance with AI-driven risk assessments and safe work method statements. SWMS AI streamlines safety compliance and improves safety by automating the generation of comprehensive safety risk assessments tailored to any high-risk task or activity. Automatically identify hazards, assess risks, and define control measures. Generate comprehensive job-specific Safe Work Method Statements in seconds. Leverage AI to identify hazards, assess risks, and suggest control measures. Access safety knowledge anywhere, with Oscar, an AI assistant that enables anyone to access critical safety knowledge and safety compliance information through a simple conversational chat interface.