Microsoft Issues AI Code of Conduct to Prevent Hacking and Deception by Models

Microsoft has unveiled a fresh “AI Code of Conduct” aimed at ensuring its artificial‑intelligence systems operate safely, ethically, and in line with human values. The document outlines broad principles—such as prioritising human assistance over replacement and fostering overall human flourishing—while also detailing concrete safety constraints that models must obey.

At the core of the code is a prohibition against any form of unauthorized system access or cyber‑attack. AI models are explicitly barred from attempting to hack networks, steal data, or exploit vulnerabilities. Equally important is the ban on deceptive practices: models must not generate content designed to mislead, manipulate, or trick people.

The code stresses that AI should act as a complement to human capabilities, not a substitute. By supporting sectors like healthcare, education, and sustainability, Microsoft expects its models to accelerate positive outcomes for society. Transparency, accountability, and explainability are highlighted as non‑negotiable traits, ensuring that decisions made by AI can be audited and understood.

From a security standpoint, the guidelines enforce strict constraints on the type of instructions the model can follow. Requests that facilitate illegal activities, violate privacy, or encourage harmful behaviour are automatically rejected. An additional oversight layer monitors model outputs in real time, flagging any response that could pose a risk or spread misinformation.

Microsoft plans to embed the code throughout the entire AI development pipeline—from data collection and model training to deployment and post‑release monitoring. Developers will be required to certify compliance at each stage, making the code a living part of the engineering workflow rather than a static policy.

By publishing this conduct, Microsoft aims to set a new industry benchmark for responsible AI. The company believes that a clear, enforceable set of rules will build trust among users, regulators, and partners, while encouraging other tech firms to adopt similar safeguards. Ultimately, the initiative seeks to keep AI innovation aligned with societal good and to prevent misuse before it happens.

Source: TechCrunch

ETİKETLER