Artificial intelligence companies on Tuesday agreed to a set of voluntary security guidelines and reaffirmed their commitment to developing their AI technology in a way that encourages the safety of their platforms and the public.
At the AI Summit in Seoul – hosted by the governments of South Korea and the United Kingdom – 16 leading companies signed the non-binding Frontier AI Security Commitments, which aim to mitigate AI risks and, at the same time, at the same time, increase accountability and transparency.
Among the companies that agreed to the commitments are OpenAI, Google, IBM, Amazon, Meta and xAI.
The companies agreed to assess technology risks at each stage of developing a new AI system, also focusing on reducing risks associated with potential misuse of the technology. They agreed to consider the results of internal and external assessments “as appropriate” to determine the security of the system. External assessments could be conducted by independent third parties, their national governments or “other bodies their governments deem appropriate”.
The companies also agreed to establish a pre-determined limit for the maximum level of risk posed by a given AI system that they would be willing to tolerate. They agreed to clearly state how they would ensure they kept risks below that threshold and what they would do if their AI system exceeded that threshold, including “processes to further develop and implement their systems and models only if they assess that residual risks would remain below of the limits.”
“In the extreme, organizations commit not to develop or implement a model or system if mitigations cannot be applied to keep risks below limits,” reads part of the non-binding agreement.
The two-day summit followed the first meeting of its kind in November and featured several world leaders who also committed to building adequate guardrails to safely develop AI and sharing resources to “forge a common understanding of AI safety and align their work on AI research.”
For the latest news, weather, sports and streaming video, go to The Hill.
/cdn.vox-cdn.com/uploads/chorus_asset/file/25550574/adt_2.jpg?w=300&resize=300,300&ssl=1 "ADT admits security breach after hackers announce stolen data on dark web")
/cdn.vox-cdn.com/uploads/chorus_asset/file/25275361/246992_AI_at_Work_OBITS_ECarter.jpg?w=300&resize=300,300&ssl=1 "Advertising companies that make money from obituary spam")
