Your Konfer December Update

Konfer is working with a US D.O.D. branch on a CRADA (Cooperative Research and Development Agreement) project on AI Safety and Assurance. The main objective of this CRADA is to develop and analyze an AI assurance capability that combines at least two approaches for evaluating the safety of AI assets. In the context of AI, assurance helps users understand if an AI asset is operating in a trustworthy, optimal, and risk-free way.

Konfer has completed a Proof-Of-Concept to prove the power of Intel® LLMs and Intel® hardware in solving the US D.O.D CRADA project of operationalizing the two AI/ML safety standards, specifically for the following three LLM use cases:

1. Setup and execute Controls Generation: Controls Generation is a generation of regulatory controls from the two AI/ML safety standards
2. Setup and execute Questions & Answering: Answering the generated controls by matching the controls to AI/ML MDLC documentation from the US D.O.D. branch
3. Setup and execute SQL Generation: Natural language to SQL generation for user friendly compliance reporting

"Konfer, through the Proof-Of-Concept has identified a solution architecture that assures reliability of Intel® inside and the power of Konfer Agentic AI guard-railed by the Intel® ecosystem. "The complexities of meeting regulatory requirements for AI compliance are challenging for enterprises, and they can become a barrier to achieving the full potential of their Gen-AI investments in a cost-efficient manner. In collaborating with Konfer and their Agentic AI Compliance and Governance Platform, Intel® is proving that the value of Intel® AI accelerators and products, both Gaudi and x86 Xeon product line can address these challenges by enabling enterprises to deploy latest Gen AI solutions complying with evolving AI regulations while achieving a high performance and return of investment."

Arijit Bandyopadhyay

CTO of Enterprise Analytics & AI,
Head of Strategy Cloud and Enterprise
CSV Group, Intel® Corporation

At its Ignite conference, Microsoft introduced Copilot Studio, a low-code platform enabling organizations to build and customize AI agents called "Copilot agents" for business tasks. These agents can autonomously interact with business systems, execute workflows, and assist individuals or teams. Microsoft aims to position these agents as the next generation of enterprise apps, emphasizing customization and automation in business processes. Early adopters, such as McKinsey and Thomson Reuters, are already utilizing these tools, reflecting growing interest in AI automation. This launch is part of a broader trend, with competitors like Salesforce and SAP introducing similar AI-based agent platforms​.

Konfer's AI agents already accomplish some of these tasks, especially the Pages feature, where AI agents read and analyze pages of documents for users. Our Konfer Playbook is trained to do this already and generate powerful questionnaires and checklists for users to meet their compliance needs. If you are interested in some of these features, contact us today.

Donald Trump's return to office is expected to prioritize industry-friendly AI policies, focusing on light regulation to stimulate innovation and compete with China. His stance may include rolling back Biden's AI initiatives, such as the U.S. AI Safety Institute, potentially weakening safeguards for ethical and national security risks. Critics argue this could empower tech giants, exacerbating monopolies. Trump's rhetoric on "woke AI" suggests concerns over perceived ideological bias, aligning with fears of politicized technology. Experts, like MIT’s Max Tegmark, emphasize balancing innovation with safety, urging education on AI risks to ensure responsible development. The approach could centralize AI power within dominant corporations, raising equity and control issues. While Trump's AI stance emphasizes economic growth, experts urge caution to address risks and harness AI's potential for solving global challenges responsibly.

Meta has made a significant shift by extending its AI model, LLaMA, to U.S. defense and intelligence agencies for national security applications, including logistics optimization, cyber defense, and counter-terrorism funding tracking. Previously hesitant to involve its technology in military uses, Meta justified this pivot as a commitment to "responsible and ethical" applications amid global AI competition. In collaboration with major tech firms such as Amazon, Microsoft, and Oracle, Meta aims to integrate LLaMA into broader defense systems, enhancing tasks like aircraft maintenance and secure data processing. The company has also granted access to its Five Eyes allies (Canada, UK, Australia, and New Zealand) to foster international collaboration​
This decision comes amid concerns about the misuse of open-source AI models and geopolitical tensions, especially as reports indicate Chinese military use of Meta’s technologies. Critics, including potential internal dissent, point to ethical dilemmas and reputational risks, especially given ongoing controversies in global conflicts. Meta's move is also seen as a way to bolster its position against competitors like OpenAI and Google, which advocate for more controlled AI development. Despite these challenges, Meta emphasizes its role in safeguarding democratic values while expanding its influence in the rapidly evolving AI landscape.

Anthropic has introduced the Model Context Protocol (MCP), a proposed open standard to help AI assistants connect seamlessly with external data sources. The protocol, which Anthropic has open-sourced, aims to address the challenge of AI models being isolated from essential data trapped in silos or legacy systems. MCP enables developers to build two-way connections between data sources and AI-powered applications, allowing these models to retrieve and interact with data from business tools, content repositories, and application development environments. Early adopters like Block and Apollo have integrated MCP, while tools like Replit, Codeium, and Sourcegraph are adding support. To facilitate adoption, Anthropic has provided prebuilt MCP servers for platforms like Google Drive and GitHub, along with toolkits to scale the protocol across enterprises.
While Anthropic highlights the potential for MCP to create a more unified, context-aware AI ecosystem, challenges remain. Rivals such as OpenAI are pursuing proprietary, partner-focused approaches for similar capabilities, such as ChatGPT's Work with Apps feature. Additionally, the lack of benchmarks or clear performance metrics for MCP raises questions about its effectiveness and potential adoption across the AI ecosystem. Despite these hurdles, Anthropic envisions MCP as a collaborative, open-source project that could drive the future of connected and context-aware AI.