This story first appeared in Sizzling Pod Insider, The Verge’s e-newsletter about podcasting and the audio trade. Enroll right here for extra.
iHeartMedia is becoming a member of corporations like Apple, Spotify, and Verizon in proscribing worker use of OpenAI’s ChatGPT — in addition to banning the usage of the AI mannequin on firm units. Based on an inner memo obtained by Sizzling Pod, iHeartMedia CEO Bob Pittman and CFO Wealthy Bressler despatched an e mail yesterday to firm staff instructing them to not use ChatGPT as a way to stop knowledge leaks of iHeartMedia’s proprietary info. The information was first reported by RBR.
The leaders wrote that “as tempting as it’s,” iHeartMedia staff had been to not use ChatGPT or related AI instruments on firm units or to do firm work — or use any firm paperwork on such platforms. This was as a way to defend iHeart’s mental property and different confidential info, in addition to that of its companions.
Of chief concern was the prospect that ChatGPT might leak proprietary info that may be precious to iHeartMedia’s rivals. Whereas OpenAI’s chatbot is primarily educated on a set of knowledge (together with web sites, Wikipedia articles, and different on-line data) that ends in 2021, the corporate additionally shops your conversations by default and makes use of them to coach its AI techniques.
Regardless of the ChatGPT ban, iHeartMedia has strongly embraced AI previously (the broadcaster as soon as rolled out AI DJs — which led to human layoffs). The corporate is at present working by itself set of AI instruments to be deployed all through the corporate.
“It should successfully practice that AI in order that anybody — even our rivals — can use it.”
“Though AI, together with ChatGPT and different ‘conversational’ AIs, might be enormously useful and really transformative, we wish to be good about how we implement these instruments to guard ourselves, our companions, our firm’s info and our consumer knowledge. For instance, when you’re importing iHeart info to an AI platform (like ChatGPT), it would successfully practice that AI in order that anybody — even our rivals — can use it, together with all our aggressive, proprietary info,” wrote the 2 leaders within the memo.
The corporate plans to roll out a set of “iHeart-specific” AI instruments that may be designed to be used internally, versus a normal market. The AI instruments would come with safeguards that may stop iHeart’s confidential info and IP from being leaked to the general public, the memo detailed. However till that point arises, staff who want to seek the advice of ChatGPT or one other third-party AI software must undergo quite a few approvals — along with getting an OK from iHeart’s authorized and IT groups.
“Within the meantime, to make sure our safety round AI, and to make sure that we don’t hurt our manufacturers or our clients’ manufacturers, or inadvertently disclose delicate knowledge, no engagement, improvement or particular undertaking work which entails ChatGPT or different AI know-how is permitted with out specific route out of your group lead. All tasks would require an evaluation of the enterprise affect and worth of the undertaking, a plan for monitoring and evaluating and a previous documented approval from Authorized and IT,” wrote Pittman and Bressler.
iHeartMedia joins a rising listing of corporations which can be drastically proscribing or banning the usage of ChatGPT and different generative AI instruments by staff — largely as a result of issues that the chatbot might leak confidential info. Main tech corporations similar to Samsung, Apple, and Verizon have blocked the web site or made it inaccessible from company techniques as a result of fears it might leak IP. Just like iHeart, Apple is working by itself AI instruments and is cautious of any proprietary knowledge touchdown within the fingers of rivals.
OpenAI has been working to deal with these issues. Now you can decide out of sharing your chat historical past with ChatGPT, and OpenAI is engaged on a set of instruments that particularly cater to companies and gained’t practice on their knowledge by default.https://hactic.s3.us-west-2.amazonaws.com/index.html