Google has warned employees against entering sensitive information into BardAI, citing vulnerabilities inherent in its chatbot technology. Getty Images
Google is concerned that vulnerabilities in its artificial intelligence chatbots could expose confidential company information, raising questions about whether its BardAI product is suitable for commercial use.
According to an anonymous source briefed on the matter, Reuters reported Thursday that the $1.6 trillion tech giant has warned employees against entering certain types of data into advanced chatbots because it could be misused. The warning also included the company’s own entrants into the AI race.
According to the newswire, Google’s privacy notice, updated in early June, states, “Don’t include sensitive or sensitive information in your conversations with the bard.”
Researchers have found that not only can human reviewers read chats, but similar AI can reproduce the data input to train neural networks, leading to backdoor leaks.
AI race
Big tech companies are in a race to see who can develop the first killer application based on generative artificial intelligence. Microsoft affiliate OpenAI launched his ChatGPT on November 30th, giving Google’s DeepMind the first blow and taking the world by storm.
Amid the hype about generative AI’s ability to pass professional entrance exams such as the bar exam, but presenting its “hallucinations” – inaccurate or outright false information – as incontrovertible fact Already, concerns are starting to arise because of the tendency to
If security risks are also found, commercial use may be severely impacted. Google is currently rolling out BardAI in over 180 countries in his 40 languages. This includes a more expensive version for business customers who don’t ingest data into public AI models.
But if even your own chatbot can’t be trusted, and you can’t prevent your competitors from reverse-engineering your trade secrets, what will your customers ultimately do?
Google’s AI track record isn’t necessarily reassuring either.
Google’s negligent action on AI ethical issues indirectly led Elon Musk to co-found his main rival OpenAI, while Google’s own AI employees are the defense At one point, he nearly rebelled over Google’s decision to work for the Ministry of Internal Affairs.
Last year, Google fired a software employee who falsely claimed that its artificial intelligence had gained sentience. In fact, large language models simply mimic human intelligence by using probabilities to predict correct outcomes.
Finally, in a rush to catch up with OpenAI, Google rushed out a BardAI presentation in February. That included a promo video that unintentionally revealed how error-prone it was. This early example of the chatbot’s hallucinatory tendencies cost parent company Alphabet a temporary $100 billion loss in market value.
Google did not immediately respond to comments, but Reutersdid not deny the reports.
