Voice Actors reject Text and Data Mining exceptions for AI training & lack of Identity Protections

The Australian Association of Voice Actors (AAVA) is disappointed at the Productivity Commission’s consideration for a text and data mining (TDM) exception for companies developing AI and the lack of consideration for restricting the use of biometric data. A TDM would retroactively legalise the theft of Australia’s creative workers’ intellectual property, voices, image, and work, by foreign multi-nationals who pay little tax and provide minimal jobs in Australia.

AAVA President Simon Kennedy :

Not only does this move put the responsibility on Australian creatives to shoulder legal battles against multi-billion-dollar tech companies who have stolen or misused their work, it also leaves the door open for AI developers to use the biometric data of every Australian whose voice or image has been recorded to train AI applications that can digitally reproduce their likeness, create deep fakes and voice clones.

Even by the Commission’s own admission “large AI models are already being trained on unlicensed copyrighted materials. In practice this has led to the scraping of voice data to create AI-generated voices that are decimating jobs in the voice acting community.

AAVA sees the Productivity Commission’s position on TDM and lack of clarity as confirmation that the exploitation of an Australian citizen’s image, voice, or likeness is not a priority.

If the Productivity Commission does not see copyright, jobs or a person’s identity as something worth protecting and places the wants of companies such as META and Google first, the question needs to be asked, who is their version of ‘productivity’ actually helping?.

A possible TDM exception to the Copyright Act, as is considered in the interim report “Harnessing data and digital technology” would harm the earning potential of actors, voice actors, journalists, presenters, dancers, singers etc. and expose them to the exploitation of their biometric data.

 

 

Considering that companies such as META via Facebook have made it clear they would like to freely use the data of everyday Australians, including children, for training AI models, the list of affected Australians will reach far beyond the creative sector.

The Australian Association of Voice Actors stands firm that AI-Training should be done on an ask-first “opt in” basis. Structures for the licensing of work and compensation to creators for the use of their intellectual property currently exist. These systems should be a minimum consideration when copyright material is being used to create a product that directly competes with an original work.

AAVA President Simon Kennedy:

Right now the balance of power lies with multi-billion-dollar tech companies, this is unacceptable. Every Australian should have to right to not participate in the training of an AI-model, especially when it comes to their own livelihood or identity, and if they choose to do so, they should be compensated fairly.

---

For more information click HERE