Tag

#safeguards

Every item tagged safeguards, newest first.

2 items

Our approach to age prediction

OpenAI is implementing age prediction in ChatGPT to identify accounts under or over 18, enhancing safeguards for teen users and iteratively improving accuracy. This aims to align with age-appropriate content policies. You can expect more granular content filtering based on predicted age. The approach will evolve with user feedback and real-world usage.

Key takeaways

Age prediction launching in ChatGPT for under/over 18 estimation.
Safeguards for teen users being enhanced.
Accuracy to improve over time with user feedback.

OOpenAI#content-moderation #age-prediction #safeguards

otherNov 7

Understanding prompt injections: a frontier security challenge

Prompt injections are a security challenge in AI systems where attackers manipulate prompts to alter model behavior. OpenAI is researching and building safeguards against these attacks. Builders should understand prompt injection risks to secure their AI applications. OpenAI's efforts aim to protect users from potential exploits.

Key takeaways

Prompt injections allow attackers to manipulate AI model behavior via crafted prompts.
OpenAI is actively researching and mitigating these security risks.
Builders must consider prompt injection risks when securing AI applications.

OOpenAI#security #frontier-ai #safeguards