OpenAI’s text-to-video AI tool, Sora, will mark a significant leap in digital content creation, opening the way for incredible opportunities and significant privacy challenges.
Unveiled on Feb. 15, OpenAI’s Sora has been the subject of much excitement. It has also sparked deep concerns among content creators and users.
One of the most significant privacy risks of Sora is that it can be used to create deepfakes — an image or video that uses a person’s likeness, usually maliciously, without their consent.
Until now, cybercriminals have had to rely on rudimentary AI video generators to create deepfakes. While these digitally altered pictures and videos can be incredibly realistic, the signs of tampering are mostly evident to the trained eye. Sora could change that.
Deepfakes and Misinformation, Privacy and Consent
Sora’s capacity to generate hyper-realistic videos from text makes it perfect for deepfakes. And, in the age of social media, these convincingly altered videos can cause untold damage to individuals, communities, and entire countries.
In political contexts, deepfake videos could be used to fabricate speeches or misrepresent public figures, influencing public opinion or even election outcomes.
The ability to generate realistic videos featuring particular individuals could also lead to grave privacy violations and further complicate the already challenging issue of consent in the digital age.
Non-consensual deepfake pornographic content featuring public figures is already a major issue that social media platforms and regulators are grappling with.
The difficulty in distinguishing between real and AI-generated content could ultimately erode public trust in visual media, complicating efforts to combat misinformation and uphold factual integrity in journalism and online discourse.
It also raises legal and ethical dilemmas, especially if such content is used in defamatory, harassing, or otherwise harmful.
Sora’s Impact on Jobs
Sora will create a major disruption in the creative sector, particularly videography, photography, and graphic design. There’s a sense of existential dread among some creatives, who question the future role of human artists in a world where AI can replicate their work effortlessly.
Some fear Sora and other AI tools will ultimately devalue traditional skills. As AI-generated content becomes more prevalent, distinguishing between authentic human creations and those generated by AI could become increasingly challenging.
A flood of AI-generated content will also raise questions about the value and uniqueness of traditional videos.
In a conversation at Davos in January, Sam Altman, OpenAI’s CEO, discussed the progression of ChatGPT models, mentioning how each successive version has become more capable, from GPT-2 being relatively limited in its capabilities to GPT-3 and GPT-4 showing linear improvements.
This means Sora’s already incredible capabilities will only become better.
How to Use Sora and Other AI Tools Safely
While OpenAI is taking significant steps to address the potential risks associated with Sora by engaging in rigorous safety measures such as red team testing, developing detection tools for misleading content, and incorporating advanced image classifiers to adhere to usage policies, the broader concerns surrounding deepfakes, misinformation, privacy, and the impact on jobs still stand.
As Sora and other AI tools become popular, it’s also important to note that many AI models are trained on public data, including the information in your queries. It’s important to avoid sharing confidential or sensitive information with AI tools.
Worried your data could be used to train AI models? You can use a data removal tool like Incogni to wipe your data from the databases of various companies. This reduces the chances that your personal information will end up in the hands of AI companies.
Read our guide to the privacy risks of AI for some actionable safety tips.
Here is a video unveiling OpenAI’s Sora:
NOTE:
Sora has no official release date yet.
For more news, follow us on X (Twitter), Threads, and Mastodon!
