Transparency, accountability 🙈

Several transparency and accountability limitations have been associated with Grok.

• Training data access. While Grok's base model has been open-sourced, the data used for training is not publicly available. This lack of transparency regarding the training dataset makes it difficult to assess potential biases or issues in the model.

• Content moderation policies. Grok is designed to answer "spicy" questions that other AI systems might reject, but the exact boundaries and moderation policies are not clearly defined or publicly available.

• Decision-making processes. There is a lack of detailed information on how Grok arrives at its responses, making it difficult for users to understand the reasoning behind its outputs.

• Third-party auditing. There is no indication of regular, independent audits of Grok's performance, biases, or potential harms, which could enhance accountability.

• User feedback. While user feedback is mentioned as part of Grok's development process, the specific mechanisms for incorporating this feedback and improving the system are not transparent.

Risks, harms 🛑

Grok has been associated with numerous risks and harms, including: 

• Inaccuracy. Grok is known regularly to "hallucinate" false information

• Misinformation. Grok regularly creates and amplifies misinformation and disinformation

• Safety. Grok generates biased, offensive and dangerous information

• Privacy. Violations of user privacy

• Environment. Grok consumes huge amounts of water and energy, contributing to climate change.

Research, advocacy 🧮

• Search Google Scholar