**The Vitality of Data in the Age of AI**
**Data-Driven Considerations in the Business World**
As artificial intelligence (AI) rapidly spreads throughout the business world, the importance of the data that fuels AI is increasing. Businesses need to consider a wide range of data-related AI considerations, including privacy, security, ethics, and training bias. Even if companies have not officially adopted AI technology, they must rethink their policies and practices related to data.
**Unrealized Internal Data Usage**
Many companies are unaware of how much internal data is already being used within their organization for AI tools. During the Fortune Brainstorm Tech conference, Dorit Dor, the Chief Technology Officer of Check Point, emphasized that the data and information of businesses may already be out there, even if they have not intentionally adopted AI technology.
**Data Leakage and Security**
As employees experiment with AI tools like ChatGPT, they unknowingly provide internal data to these systems. This presents significant challenges in terms of data leakage, which can compromise proprietary competitive information and personal customer data. Clara Shih, CEO of Salesforce’s AI business, highlighted the lack of a clean separation between secure databases and AI tools. She explained that the large language models powering generative AI tools require as much context as possible from users to produce relevant and accurate responses. Failure to architect these tools carefully can result in the model learning and storing sensitive data.
**Ensuring Data Protection through Security Best Practices**
Sean Scott, the Chief Product Development Officer of PagerDuty, emphasized the importance of following security best practices to address data leakage concerns. He highlighted the need for companies to establish clear policies regarding data protection and educate employees about these policies. Monitoring and enforcement mechanisms should also be in place to ensure policy adherence.
**Safeguarding against Mystery Model Data**
Aside from protecting internal data, companies must also address the quality of external data they ingest into AI models. Signal President Meredith Whittaker warned that most off-the-shelf AI large language models are black boxes, making it difficult for companies to know the characteristics of the data used for training these models. Implementing such AI tools without a clear understanding of the underlying data can result in incorrect or offensive results.
**Fine-Tuning AI Models and the Need for Regulation**
Whittaker emphasized the need for additional regulation to address the lack of transparency surrounding the data used in AI models. She suggested that fine-tuning these models with additional data could mitigate drawbacks, such as offensive results. However, she called for clearer guidelines and limitations on the data used to ensure ethical and accurate outcomes. While regulation can raise the minimum requirement for data usage, Dorit Dor cautioned that it is not sufficient to attain complete safety in the AI space.
**The Burden on Chief Information Security Officers**
The responsibility of managing data in the AI era falls heavily on the shoulders of Chief Information Security Officers (CISOs) within companies. Dorit Dor acknowledged that CISOs were already overwhelmed with their existing responsibilities, and now they must navigate new challenges related to AI and data, including legal aspects.
In conclusion, as AI becomes more prevalent in business operations, companies must prioritize data security, address the challenges of data leakage, and ensure the quality of external data used in AI models. Clear policies, employee education, and monitoring mechanisms are essential for protecting internal data. Additionally, regulatory interventions and guidelines are necessary to improve transparency and prevent the use of problematic data. The role of CISOs in managing data-related issues in the AI era should not be underestimated.
GIPHY App Key not set. Please check settings