OpenAI has presented a safety framework for its advanced AI models, allowing the board to overturn safety decisions as part of the plan released on its website. Supported by Microsoft, OpenAI will only deploy its latest tech if it’s deemed safe in key areas like cybersecurity and nuclear threats. They’re also establishing an advisory group to review safety reports, forwarding them to company executives and the board. While executives hold decision-making power, the board retains the authority to reverse these decisions.
The framework outlines strategies to monitor, evaluate, and safeguard against potential risks associated with increasingly powerful AI. It integrates various safety teams to mitigate current risks, anticipate emerging threats, and establish safety protocols. Employing rigorous assessments and continuous updates, OpenAI aims to ensure the safe development and deployment of frontier AI models. This evolving framework emphasises ongoing improvement, collaboration, and external accountability.
Since ChatGPT’s launch, concerns about AI risks have been prominent among both researchers and the public. While generative AI has impressed users with its creative abilities, it has also raised worries due to its potential for spreading disinformation and manipulating individuals.
Why does this matter?
In April, a coalition of AI experts called for a halt in developing systems more powerful than OpenAI’s GPT-4, citing societal risks. A May Reuters/Ipsos poll discovered that over two-thirds of Americans harbor concerns about AI’s potential negative impacts, with 61% fearing it could pose a threat to civilization. Furthermore, the steps comes after the conflict faced between CEO Sam Altman and the board of OpenAI over safety questions, where Altman was fired and re-hired and the OpenAI board faced some changes.
The Biden administration has taken initial actions to establish essential guidelines ensuring the safe utilization of generative artificial intelligence (AI) systems. The National Institute of Standards and Technology (NIST) under the Commerce Department seeks public input to develop pivotal standards and testing protocols for securing AI systems by February 2, 2024. This initiative, driven by President Biden’s AI-focused executive order, aims to cultivate industry standards prioritizing safety, security, and trust in AI technologies.
As reported by Reuters, NIST is formulating assessment criteria for AI, nurturing standards, and creating test environments for AI systems. The request for input encompasses the management of risks linked to generative AI and curbing the spread of AI-generated misinformation. Additionally, NIST is developing testing guidelines, focusing on the strategic application of “red-teaming” for effective AI risk assessment and management, and establishing corresponding best practices.
Pope Francis has called for a global agreement to control artificial intelligence (AI), stressing the importance of safeguarding human principles amid technological progress. In his address marking the World Day of Peace, he cautioned against the potential supremacy of algorithms, calling for international collaboration in supervising the development and utilisation of AI.
The Pontiff stressed the need for ethical examination, expressing worries about AI applications that could imperil humanity and the environment. Specifically, he raised concerns about AI’s involvement in weaponry, emphasising the irreplaceable role of human ethical judgement in decisions concerning autonomous lethal systems. While acknowledging the benefits of technological advancements, Francis highlighted the pivotal role of AI regulation, deeming it a crucial consideration for the future of humanity. His appeal aligned with global endeavours to navigate the pros and cons of AI, coinciding with the recent negotiation of significant AI regulations by the European Union.
Why does this matter?
Pope Francis joins an array of public figures and institutions calling for ethical AI regulation and implementation. This comes amidst new AI regulations being discussed and launched across the globe.
On December 18, 2023, ISO/IEC 42001 on artificial intelligence (AI) management systems was published to assist organizations in developing a strong AI governance framework.
ISO/IEC 42001 is ‘an international standard that provides a framework for establishing, implementing, maintaining, and continually improving an Artificial Intelligence Management System (AIMS).’ The standard covers key components such as transparency, explainability, and autonomy, and it includes various requirements for managing AI systems effectively. These provisions cover ‘leadership, planning, support, operation, performance evaluation, and continual improvement.’
When deploying AI systems, ISO/IEC 42001 highlights ethical principles and values such as fairness, non-discrimination, and respect for privacy. ISO/IEC 42001 is part of a broader set of standards aimed at governing best practices for trustworthy AI deployment and improvement. It serves as the foundation for external certification and auditing of AI systems in accordance with the risk assessment methodology in the future ISO/IEC 42006 standard. ISO/IEC 42001 is scalable, making it relevant for organizations of all sizes and sectors.
Why does it matter?
ISO/IEC 42001 is the first international management system standard for the safe and reliable development and implementation of AI, aiming to help businesses and organizations develop a robust AI governance framework. The standard is applicable to all types of companies in any industry and is the only AIMS that is certifiable. It is comparable to ISO 9001 on quality management and ISO 27001 on information security, providing best practices, rules, definitions, and guidance to manage risks and operational aspects of AI systems. The new ISO standard is designed to promote the development and use of AI systems that are trustworthy, transparent, and accountable. It will help organizations identify and mitigate risks related to AI development and implementation, prioritize human well-being, safety, and user experience in AI design and deployment, and guide organizations in complying with relevant legislation, regulations, data protection rules, or obligations towards stakeholders.
To address some of the worries about AI, governments throughout the world are racing to propose rules and regulations to govern its usage, such as in the US with the Biden executive order on safe, secure, and trustworthy development and use of AI and the European Union’s AI Act. The new ISO/IEC standard specifies the requirements for a certifiable AI management system framework, allowing organisations to maximise advantages while assuring stakeholders that systems have been established and are being managed ethically. It provides a comprehensive framework for managing AI systems, enabling global interoperability, and laying the groundwork for responsible AI development and implementation. It is an essential tool made available to organizations to enable ethical and trustworthy AI model development and deployment.
The European Union has reached a historic agreement on the world’s first comprehensive set of rules to regulate AI. The legislation, known as the AI Act, sets a new global benchmark for countries seeking to harness the potential benefits of AI while trying to protect against its possible risks.
The main new elements of the provisional agreement can be summarised as follows:
General purpose AI (GPAI) systems: The agreement specifies that GPAI must follow the transparency criteria outlined by Parliament. These encompass creating technical documents, abiding by EU copyright regulations, and distributing comprehensive overviews of the content utilised for training these models.
When it comes to the high-impact GPAI models with systemic risk, they will face extensive obligations, including evaluation, risk assessment, cybersecurity, and energy consumption reporting. At the same time, lawmakers agreed that the codes of conduct would serve as supplementary guidelines for GPAIs with systemic risk until technical standards are harmonised.
High-Risk Use Cases: A consensus was reached on the roster of high-risk AI applications, with a focus on safeguarding individuals’ safety and fundamental rights, with a mandatory fundamental rights impact assessment. Citizens will have the right to fail complaints regarding AI systems and to receive clarifications about decisions influenced by high-risk AI systems that affect their rights.
Governance: An EU AI office will be established within the commission to enforce foundational model rules, with national authorities overseeing AI systems through the European Artificial Intelligence Board (EAIB) for consistent application of the law. An advisory forum will gather feedback from stakeholders. A scientific panel of independent experts will advise on enforcement, identify systemic risks, and contribute to classifying AI models.
Biometric categorisation systems that use sensitive characteristics ( including political, religious, philosophical beliefs, sexual orientation, and race),
Untargeted scraping of facial images from the internet or CCTV footage to create facial recognition databases,
Emotion recognition in the workplace and educational institutions,
Social scoring based on social behaviour or personal characteristics;
AI systems that manipulate human behaviour to circumvent their free will, and
AI used to exploit the vulnerabilities of people (due to their age, disability, social or economic situation).
Law Enforcement Exceptions: Narrow law enforcement exceptions were introduced in terms of real-time remote biometric identification (RBI). These exceptions will be subject to judicial authorisation primarily and for strictly defined lists of crimes including:
Targeted searches of victims (abduction, trafficking, sexual exploitation),
Preventing terrorism, and
The localisation or identification of a person suspected of having committed one of the specific crimes mentioned in the regulation such as terrorism, trafficking, sexual exploitation, murder, kidnapping, rape, armed robbery, participation in a criminal organisation, and environmental crimes.
The AI Act is expected to come into force no earlier than 2025, requiring EU countries to formalise or create a national AI regulator on top of a European watchdog.
Why does it matter?
The legislation needs to go through a few final steps for final endorsement, but the political agreement means its key elements have been approved. The AI Act is a significant step in the EU’s efforts to bring a new level of oversight to tech, akin to regulation of the healthcare or banking industries. With the EU being the first to set rules on AI, this legislation could represent a standard and a model to follow for many other jurisdictions around the world, which could fuel another Brussels effect.
The UNCTAD eWeek 2023 session ‘Bottom-up AI and the right to be humanly imperfect.‘ offered an in-depth examination of the far-reaching impact of AI, spanning its transformative potential across industries and the intricate dynamics entwined with its integration into education and political decision-making processes. Discussions within the panel unveiled insights that emphasized the significance of transparency, inclusivity, and a measured approach to integration, unveiling the complex tapestry of AI technology and its interface with evolving political landscapes.
Jovan Kurbalija highlighted the remarkable ability of AI to discern complex patterns within vast datasets, leading to accurate predictions across diverse sectors like healthcare, finance, and transportation. He emphasized the necessity of avoiding monopolization in AI development by fostering open-source models, promoting collaboration, and preventing the stifling of innovation by large corporations.
Ownership and provenance of AI knowledge
Kurbalija stressed the importance of clearly attributing AI knowledge sources, advocating for transparency and shared ownership. He illustrated this with Diplo’s utilization of AI for teaching, emphasizing the need to acknowledge AI knowledge sources to maintain fairness and transparency.
AI in education: Balancing benefits and concerns
While highlighting AI’s positive impact on individual learning by aiding in understanding complex subjects, Kurbalija cautioned against the potential ban of AI in education. Instead, he advocated leveraging AI as a valuable tool to enhance the learning process, urging the promotion of critical thinking within academic institutions alongside its use.
Challenges in political decision-making with AI integration
Kurbalija cautioned against excessive computerization of political processes, emphasizing the inherently human aspect of politics. He highlighted the need for careful integration of AI as a tool in political decision-making to maintain integrity and preserve the human touch in these processes.
Outsourcing information and institutional thinking
Yung-Hsuan Wu deliberated on the debate surrounding the outsourcing of crucial information to third-party tools. While acknowledging the efficiency of such tools, concerns were raised about the potential loss of control over interpretation and the emphasis placed on certain details. The discussion underscored the significance of institutional thinking, highlighting its role in strengthening decision-making within institutes.
Just-in-time reporting from the UNCTAD eWeek 2023
Stay tuned for just-in-time reporting facilitated by The Digital Watch Observatory’s AI-driven App! Keep this page bookmarked to access comprehensive reports from all special, thematic and high-level sessions. Please note that session summaries will promptly appear on our dedicated page following their conclusion.
Download our free mobile app and read UNCTAD eWeek 2023 reports on the go!
Download the Dig.Watch News+mobile app from Google Play or the App Store to have instant access to our reports on your mobile device.
The Council has established its position, known as the ‘general approach,’ regarding the Artificial Intelligence Act within the EU. This act is geared towards ensuring the safe and lawful usage of AI systems in the EU market while upholding fundamental rights and the values of the Union. The proposed regulation, a pivotal component of the EU’s AI policy, takes a risk-based approach, providing a standardized legal framework to ensure adherence to regulations.
Key elements encompass:
Clearly defining AI systems, with a focus on machine learning and logic-based methods.
Prohibiting specific AI practices, like social scoring and the exploitation of vulnerable groups.
Categorizing high-risk AI systems and outlining detailed compliance requirements.
Clarifying responsibilities along AI value chains and aligning them with existing laws.
Addressing the integration of general-purpose AI systems into high-risk ones.
Excluding national security, defense, and non-professional use from the Act’s scope.
Improving the compliance framework, streamlining market surveillance, and empowering the AI Board.
Applying proportionate penalties for non-compliance, considering SMEs and startups.
Prioritizing transparency, such as public entity registration and user alerts for emotion recognition systems.
Introducing measures that encourage innovation-friendly regulations, like AI regulatory sandboxes and provisions for real-world testing.
Why does this matter?
The Council’s adoption of this general approach initiates negotiations (‘trilogues’) with the European Parliament, aiming for a final consensus on the proposed regulation. This underscores the need to balance innovation with safeguarding fundamental rights in utilising AI within the EU market.
The battle for leadership at OpenAI has taken a dramatic turn with the dismissal of Sam Altman, the CEO, by the company’s board. Microsoft, one of the main founders of OpenAI, is now putting pressure on the board to reinstate Altman. The conflict at OpenAI revolves around two co-founders: Altman, who is focused on business opportunities, and Ilya Sutskever, who is concerned about the security risks posed by OpenAI’s technology.
Under Altman’s leadership, OpenAI experienced significant success, particularly with its popular ChatGPT chatbot. Altman became a prominent figure in the tech industry as a result. However, tensions arose within the company as Sutskever, an esteemed AI researcher and board member, grew increasingly concerned about the potential dangers of OpenAI’s technology. He also voiced discontent about his diminished role in the company.
When Sutskever led four out of the six board members in abruptly removing Altman from his position, the conflict between rapid growth and AI safety had reached a critical point. This unexpected move shocked OpenAI employees and the wider tech industry, including Microsoft, which has invested $13 billion in the company. Some industry insiders compared this split to Steve Jobs’ departure from Apple in 1985.
What is the reason for firing Altman? Besides mentioning communication issues in a blog post, the board has not given any specific justifications for dismissing Altman. OpenAI employees were assured that the decision was unrelated to financial, business, safety, or security/privacy concerns.
‘Mr. Altman’s departure follows a deliberative review process by the board, which concluded that he was not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities,’… ‘The board no longer has confidence in his ability to continue leading OpenAI.’ The company explained in a blog post.
Shortly after leaving his position speculation arose regarding Altman’s next steps. Initially, there were indications that he might rejoin OpenAI, followed by rumors of establishing a new venture with close associates. However, the most recent updates suggest that Altman is likely to join Microsoft.
The aftermath of Altman’s dismissal has plunged OpenAI into turmoil, with co-founder Greg Brockman and director of research Jakub Pachocki resigning in protest, along with other key figures. OpenAI’s approximately 700 employees are left bewildered, trying to make sense of the board’s decision.
Moreover, the backgrounds of some board members provide insight into their motivations. Board members with ties to the Rationalist and Effective Altruist movements, concerned about the potential threats of AI, suggest that philosophical and ethical considerations played a significant role in the board’s choices.
Altman transformed OpenAI into a for-profit company in 2018, securing substantial investments from Microsoft. The company’s success, coupled with the release of technologies like GPT-4, led OpenAI to seek further funding and explore the development of hardware devices for AI technologies.
In conclusion, the leadership changes at OpenAI highlight the challenges of balancing business opportunities with security risks in the rapidly evolving AI field. They also underscore the deep divisions within the AI community regarding AI development’s pace and potential dangers. OpenAI must navigate these concerns while striving to maintain its success in the industry.
The recent disruption of OpenAI’s ChatGPT service, occurred on 8 November, has sparked waves of concern within the AI and natural language processing communities. With more than 100 million active weekly users, ChatGPT, a stalwart in the AI landscape, faced an unexpected blackout lasting over 90 minutes. However, the repercussions extended beyond ChatGPT, casting a shadow over OpenAI’s entire ecosystem as its API services also succumbed to the disruption.
The ChatGPT and associated services disruption has garnered significant attention not only due to its duration and extent, but also because it marked the second outage within a 48 hours timeframe. A partial outage had occurred on Tuesday, 7 November, following the one on Thursday, 19 October and the previous one of Friday, 15 September. These partial outages, together with the last, major one, raise concerns not only about the service’s stability and reliability, but also about the impact future outages may have on global business development.
OpenAI, a trailblazing company at the forefront of AI innovation, has been actively addressing this issue. They’ve implemented a fix and brought ChatGPT service back to its duties, assuring users of their commitment to resolving the problem swiftly and effectively. However, according to the last updates from OpenAI status website, it’s been stated that the company is ‘dealing with periodic outages due to an abnormal traffic pattern reflective of a DDoS attack’ and is continuing to work in order to mitigate similar issues.
Presumed cyberattack
In a recent development, the hacktivist group, Anonymous Sudan, has claimed responsibility for the cyberattack on OpenAI’s ChatGPT. The group outlined its motives in a post on their Telegram channel. The primary reasons cited for targeting ChatGPT include OpenAI’s collaborations with Israel, particularly emphasising the CEO’s expressed intention to increase investments in the country.
Anonymous Sudan underlined its focus on American companies as a driving force behind the attack. Additionally, the group claimed an alleged bias in ChatGPT, suggesting a preference for Israel over Palestine within the chatbot’s interactions. These stated reasons shed light on the complex motivations behind the cyber assault, intertwining geopolitical concerns and perceived biases within the AI realms.
Wide-scale adoption of AI services
These unexpected outages brought many users to question the matter of sustainability of such backing infrastructure, since many companies are increasingly adopting ChatGPT’s services and counting on its support to discharge their daily tasks. The trajectory of AI adoption within a wide range of industry sectors is poised for significant growth from 2022 to 2025. As of 2022, nearly half of the executives surveyed had the expectation that their respective companies would undergo wide-scale adoption of AI technologies. This anticipation underscores the recognition of AI as a transformative force within the world economy, with executives foreseeing the integration of AI across various facets of their operations.
Looking ahead to 2025, the sentiment among these industry leaders is even more optimistic. There is an expectation that the adoption rate of AI will not only continue to grow, but will surpass the earlier projections for wide-scale implementation. Executives envision a future where AI becomes not just a supplementary tool, but a critical component deeply embedded in the operational fabric of their companies.
This shift in expectations from wide-scale to critical implementation reflects a growing understanding of the profound impact that AI can have on the global economy workflow. AI is increasingly seen not merely as a trend or optional enhancement, but as an integral element that can drive efficiency, innovation, and strategic decision-making.
The factors driving this anticipated surge in AI adoption within the global work environment are manifold. The promise of enhanced data analysis, streamlined processes, and the ability to derive actionable insights from vast datasets positions AI as a catalyst for operational excellence. Additionally, as AI technologies continue to mature and demonstrate tangible benefits, companies are more inclined to invest in and fully embrace these advancements.
The increasing adoption of AI tools into various aspects of personal and professional lives is sparking a question about whether it will be possible to work and live without their backing in the near future, considering that once we got used to them, it will not be easy to replace automated processes and thus overcome longer inaccessibilities and misfunctions. Analysing this scenario requires consideration of the current role of AI, its impact on different sectors, and potential future developments, as the recent outage of OpenAI’s services highlighted the need for stronger infrastructure and consistent service reliability.
Knowledge slaveries: Is Bottom-up AI a possible solution?
The constant use of AI services reveals our thoughts and emotions through interactions with AI platforms, resulting in a vast amount of data that can be used to extract patterns in our thinking. This trend has given rise to a new AI economy where these patterns are collected, codified, and monetised, raising concerns about privacy and cognition intrusions beyond what social media and tech platforms currently pose.
This development risks creating a state of ‘knowledge slavery’, where corporate or government AI monopolies control access to our knowledge. To counter this, it is essential to retain ownership over our thinking patterns, including those derived automatically through AI.
One of possible solution lies in the development of bottom-up AI, Diplo’s Executive Director Dr Jovan Kurbalija claims. Bottom-up AI is both technically feasible and ethically desirable, and has the potential to address governance concerns raised by generative AI tools like ChatGPT. It gives control back to individuals and communities, ensuring privacy and data protection. It also fosters inclusivity, innovation, and democracy by mitigating the risks of power centralisation inherent in generative AI.
Contrary to the prevailing belief that powerful AI platforms can only be built using big data, leaked documents from Google suggest that open-source AI could outperform proprietary models like ChatGPT. Open-source platforms such as Vicuna, Alpaca, and LLama are already offering similar quality, while being more cost-effective, faster, more modular, and greener in terms of energy consumption.
The technology for bottom-up AI is advancing, but there is a need to ensure the quality of data. Currently, data labelling is mainly performed manually in low-cost English-speaking countries, risking labour law and data protection challenges. Diplo, a leading organisation, integrates data labelling into their daily operations, gradually building bottom-up AI by digitally annotating text during research and other tasks.
While the full adoption of bottom-up AI remains uncertain, it may coexist with top-down AI approaches. Some individuals and communities may be more inclined to experiment with and embrace bottom-up AI, while others stick to top-down AI due to inertia. However, questioning the prevailing AI paradigm and exploring alternatives is crucial to make informed decisions that benefit society as a whole and to prevent inconveniences from future outages of bigger AI service providers people are relying on.
On 1 November 2023, the First Committee (Disarmament and International Security) of the UN General Assembly approved a draft resolution on lethal autonomous weapons systems (LAWS), expressing concern about the possible negative consequences and impact of autonomous weapons systems on global security and regional and international stability and stressing the urgent need for the international community to address the challenges and concerns raised by such systems.
The resolution, once endorsed by the General Assembly, would require the UN Secretary-General to seek the views of Member States and observer States on LAWS and on ways to address the challenges and concerns they raise from humanitarian, legal, security, technological, and ethical perspectives, and to submit a report to the General Assembly at its seventy-ninth session. The Assembly would also request the Secretary-General to invite the views of international and regional organizations, the International Committee of the Red Cross, civil society, the scientific community and industry and to include those in the annex to the report.
Within the First Committee, the draft resolution was adopted by a vote of 164 in favour to 5 against (Belarus, India, Mali, Niger, Russian Federation), with 8 abstentions (China, Democratic People’s Republic of Korea, Iran, Israel, Saudi Arabia, Syria, Türkiye, United Arab Emirates). In addition, 11 votes were recorded on the resolution’s provisions.
Egypt noted that algorithms must not be in full control of decisions that involve harming or killing humans. Human responsibility and accountability for the use of lethal force must be preserved.
The Russian Federation expressed concern that the resolution seeks to undermine the work of the GGE on LAWS, which is the sole ideal forum to discuss LAWS. The country also argued that the resolution does not acknowledge that autonomous weapons systems can play an important role in defence and in fighting terrorism, and that international law fully applies to these systems.
Iran noted that the definition and scope of the term ‘lethal autonomous weapons’ are not clearly defined, and that GGE on LAWS should focus on states parties.
Türkiye also raised the issue of a lack of agreement on the definition of autonomous weapons systems and noted that the absence of shared terminology increases ‘question marks’ on the way forward. The country also added that international law and international humanitarian law should be sufficient to alleviate concerns regarding the use of such weapons systems.
The USA stated that it does not support the creation of a parallel process on LAWS or any other efforts that will seek to undermine the centrality of the GGE on LAWS on making progress on this issue. Poland also noted that the GGE is the forum to make progress on identifying challenges and opportunities related to LAWS, and that other international forums are not equally fit, as they often lack technical and diplomatic capacity and do not address the significant balance between humanitarian aspects and military necessity.
Israel called on member states not to undermine the work done in the Convention through the creation of a parallel forum. It also outlined the importance of the full application of international humanitarian law to LAWS.
Australia called for the report to be prepared by the UN Secretary-General to be balanced and inclusive of the views of all UN member states. South Africa expressed concern about the provision of the resolution, noting that the integrity of the process under way in the GGE on LAWS should be respected, and states parties have already made their views known on the issue. Brazil argued that the GGE might benefit from the fresher views of a wider audience.
