ChatGPT with Care: A Corporate Guide
Think of ChatGPT as the versatile new hire who’s ready to lend a hand with just about anything. It’s streamlining tasks for everyone, from developers who are stuck fixing annoying bugs to Project Managers having an easier time streamlining project timelines. Analysts can quickly whip up data interpretations, while the higher-ups get a clear view of the market landscape.
….But as with everything, it has its own pitfalls. ChatGPT can sometimes let slip confidential info if it’s not managed right. The headlines have done their rounds — tales of when ChatGPT got too chatty and companies had to slam the brakes on it to keep things under wraps.
These incidents bring light to several dangers that using ChatGPT entails:
Misinformation:
- Inaccuracies : ChatGPT sometimes tends to hallucinate and provide inaccurate or misleading information which could be detrimental if used in critical decision-making scenarios².
- Misinterpretation: Due to its lack of understanding, ChatGPT might misinterpret user input leading to incorrect or irrelevant responses which could mislead users.
Privacy Issues:
- Sensitive Data Exposure: Sharing sensitive business data with ChatGPT could potentially expose companies to risks, if the data is mishandled or misused³.
- Data Privacy Laws: Businesses can run afoul of privacy laws when using ChatGPT, especially if it leads to unauthorised data handling or sharing⁴
Dependency:
- Over-Reliance: The ease of use and the broad range of assistance provided by ChatGPT could lead to an over-reliance on the tool, which might hinder the development of critical thinking and problem-solving skills among IT professionals
- Automated Error Propagation: As ChatGPT and similar AI tools take on more automated tasks, there’s a risk of propagating errors at a larger scale, especially if the AI generates incorrect or sub-optimal solutions which get implemented without thorough review.
Industry giant’s response to these challenges
Given the huge risk to the companies and the clients, the giants have taken the blunt way of “solving” the issue by restricting ChatGPT usage for all their employees.
A More Nuanced Approach:
Banning ChatGPT might have been a necessary temporary halt, but it is definitely not the solution going forward.
As mentioned in these⁵⁶ studies, the use of ChatGPT can lead up to to 36% increase in efficiency in white collar jobs. This technology is here to stay, and its potential to revolutionise multiple industries is significant. Shying away from it is not an option.
Moreover, banning a particular tool like ChatGPT is futile, as a quick search reveals numerous alternatives⁷ with similar functionality and risks.
The viable way forward is through education and cautious utilisation. As a wise man once said:
Best Practices for Responsible ChatGPT Usage
Opting out of data collection
The simplest and the most effective step is to opt out of data collection.
Within the Settings tab of the ChatGPT UI, they provide a way to opt-out of collecting your data .…..but
It also turns off the chat history, which disables us from retrieving or continuing any old conversations. On top of that, it doesn’t sync across devices, so you’ll have to turn this off on all devices.
Both of those are quite annoying. This is just OpenAI’s sly way to keep users from turning it off.
But no need to worry, there is another not-so-obvious way to turn data-collection off as well.
You can do this by going to the new Privacy Portal of OpenAI and requesting to not train on your content.
You agree to the terms and fill in your location and then confirm your request.
This is a persistent way of turning it off and you will still get to keep your Chat History 🎉🎉
Requesting deletion of personal data
In case you do find your personal information on ChatGPT, you can request OpenAI to remove it through this form.
Compliance Checklist
Prior to using ChatGPT, create a checklist to assess the sensitivity of the data in your query to prevent potential leaks.
A brief checklist for example:
- Data Classification: Identify if the data is sensitive, proprietary, or general.
- Data Segregation: Separate sensitive or proprietary data from general data before sharing with ChatGPT.
- Verification: Have reliable sources ready for cross-verifying ChatGPT’s responses.
- Compliance: Adhere to organizational and legal data handling guidelines.
Enterprise edition
If possible and feasible, try to get the enterprise version of ChatGPT. This offers better compliance with data protection regulations like GDPR or CCPA, and provides clearer guidelines regarding ownership and usage rights. It also adds extra encryption on top. You can learn more about it here.
Custom models
If you want complete control over your data, you can also consider deploying an LLM of your own on the cloud or in-house. There are quite a few open-source models available which are deploy-ready.
Interested in learning more about using open-source models? Stay tuned for an upcoming article where we’ll delve into the options, challenges, and essential considerations for a successful deployment.
Conclusion
ChatGPT’s dual potential for boosting efficiency and posing risks calls for a balanced corporate strategy. The key lies not in avoidance but in informed adoption and robust data management practices. By embracing a responsible approach that includes thorough data checks, opting out of unnecessary data collection, and exploring secure enterprise options, businesses can benefit from ChatGPT’s capabilities while minimizing the pitfalls of AI integration.
