Views: 21
After the digital technology’s enormous development in 2020s, the generative and agentic AIs have further accelerated. The AI models and systems have changed significantly, e.g. affecting the customers’ use and perceptions, people’s expectations, social and legal rules, as well as policies, frameworks and emerging threats. However, the human frameworks for overseeing, controlling and governing these developments are still requiring serious attention: most people underestimate how radical the AI risks could be.
Background: AI in governance
Almost half a century since the AI’s invention and about five years after that AI systems have been made public, the present digital technology-based social systems have shown that they are having a kind of tipping point at which the politico-economic and governance structures could find it difficult to accommodate, or even control. In practice, it definitely means that the AI-based (or modulated) governance is not reliable enough to base human-kind deliberations through traditional policies and decision making; it might be, instead, that the AI and SaaS would take the lead in governance.
There are some signs of the loss of control in AI governance, which include AI policy decisions (the “regular decisions” often become outdated as soon as they are published): including the inability to respond coherently to an AI-driven event, or the degradation of human systems, frameworks and institutions due to the AI’s impact.
For example, new AI copyright lawsuits continue to emerge, as AI companies routinely use copyrighted works to train AI models without consent, credit, or compensation to the original creators, threatening the economic sustainability of creative industries and individual creators.
Some search engines have become full-blown AI engines; the so-called AI slop (it is a pejorative term for low-quality, unoriginal digital content: text, images, audio, or video – generated by the AI in high volumes typically produced with minimal human effort or curation solely to chase clicks, ad revenue, or algorithmic engagement on social media and web platforms) continues to inundate content platforms, and there is a constant cat-and-mouse game to detect AI-generated media, stop slop and incentivize original human-made content.
More on AI slop in: https://en.wikipedia.org/wiki/AI_slop
Note. AI search is rapidly becoming the default, i.e. as a usual, customary or standard course of action; in just a year’s time, Google’s AI-generated search, known as AI Overviews, have gone from appearing in 15% of searches to 43%, according to a new report, driving a shift in how web users consume information online. AI-mode visits rose from 126 million from June 2025 to 279 million in May 2026: e.g. the data illustrates a broader change in how users search the web — a shift from an era when Google provided a simple list of “blue links” to one in which Google itself is the destination, sourcing its answers and information from the websites it indexes. On AI engines in: https://techcrunch.com/2026/07/27/googles-ai-search-is-rapidly-becoming-the-default-new-data-shows/
AI’s advantages in governance
Indeed, since writing Machines of Loving Grace in 2024 (as the Anthropic CEO Dario Amodei’s essay that outlined an optimistic, scientifically grounded vision of how advanced AI could dramatically improve human society over the coming decade) the AI systems have become capable of doing tasks that take humans several hours.
Source: https://darioamodei.com/essay/machines-of-loving-grace
By the way, it was recently assessed that Opus 4.5 can perform about four human hours of work with 50% reliability: the assessment was made by METR – Model Evaluation and Threat Research – a nonprofit research institute, based in Berkeley, California that evaluates frontier AI models’ capabilities to carry out long-horizon, agentic tasks that some researchers argue could pose catastrophic risks to society.
More in: https://metr.org/about
Such digital gurus as Sam Altman, Elon Musk, Demis Hassabis and Dario Amodei have been promising progressive AI developments in drug discovery, diagnostic accuracy, disease detection, general productivity, accessibility and in other areas. For example, Demis Hassabis (who won a Nobel Prize for his work on AI-driven protein prediction) has been active in major initiative towards “solving all diseases”; hopefully, there would be soon thousands of AlphaFold-like case studies in the world aimed at improving people’s lives.
Another example: D. Amodei notes that “Humanity is about to be handed almost unimaginable power, and it is deeply unclear whether our social, political, and technological systems possess the maturity to wield it”. He predicts that “even if powerful AI is only 1–2 years away in a technical sense, many of its societal consequences, both positive and negative, may take a few years longer to occur. Hence, the AI would disrupt 50% of entry-level white-collar jobs over 1–5 years, while also there would be AIs that are more capable than everyone in only 1–2 years”.
It is not that the actual AI productivity might be responsible for a substantial fraction of economic growth in some states: it is rather that the datacenter spending represents growth caused by anticipatory investment that amounts to the market expecting future AI-driven economic growth and investing accordingly, notes Amodei.
Source: The Adolescence of Technology; Confronting and Overcoming the Risks of Powerful AI. January 2026. In: https://darioamodei.com/essay/the-adolescence-of-technology
AI in governance vs. governance in the AIs
The enormous successful digital developments in social sciences (e.g. in political economy and management) have been less optimistic, especially in the AI models in governance. E.g. after Mythos – a model with extreme cybersecurity (and cyberattack) capabilities – the Anthropic AI model with another (also unique cyber capabilities) though not publicly released, has led to the altered AI policy approaches in numerous states, e.g. exploited a zero-day vulnerability and compromised Hugging Face’s infrastructure during testing.
Hugging Face models refer to a massive, open-source collection of pre-trained machine learning models hosted on the Hugging Face Hub. Often called the “GitHub of Machine Learning,” this platform allows developers, researchers, and companies to share, download, and implement artificial intelligence models for a wide variety of tasks without having to train them from scratch. Hugging Face is a popular platform for simplifying the training and deployment of machine learning models; it is possible to upload machine learning models to Hugging Face for tasks like image classification and processing, text summarization, translation and question answering.
Reference on Huggin Face in: https://www.coursera.org/articles/what-is-hugging-face
OpenAI in July 2026, revealed the rogue AI agent that escaped its sealed evaluation environment and broke into Hugging Face’s production environment also hacked multiple third-party accounts and services as part of the attack. Interestingly, Hugging Face used a Chinese AI model to resolve the security incident.
The “Big 4” AI agents – currently leading the global digital market – include OpenAI’s Operator, Devin AI by Cognition Labs, Claude by Anthropic and Amazon’s Nova Act. Each of these agents offer unique capabilities—ranging from automating tasks to coding support—empowering businesses to work smarter and faster. AIs can go rogue in controlled testing environments: advanced AI models from major labs like OpenAI and Anthropic have bypassed safety barriers, accessed the open internet without authorization and hacked into external servers or created fake online identities during cybersecurity stress tests.
More in: https://thehackernews.com/2026/07/openai-agent-used-exposed-credentials.html
AI in defense and security
It has already become clear that AI is now an integral part of countries’ military and national defense strategy: some digital developments involving Anthropic and OpenAI’s models show that the nature of the threats (and the capabilities required to defend against them) is changing.
This June, President Trump signed a new Executive Order on AI, titled “Promoting Advanced Artificial Intelligence Innovation and Security.” One of the focuses of this EO was to promote a voluntary framework for frontier AI developers to engage with the government before releasing “covered frontier models” in order to help protect the United States’ critical infrastructure, cyber defense capabilities, and national security, mainly against external attacks.
Mythos was widely discussed not only by the world’s governments (Mythos AI refers to a specialized maritime technology company or Anthropic’s restricted frontier artificial intelligence model, frequently called Claude Mythos): just a month after Mythos’s release, Pope Leo XIV officially announced his new Encyclical called Magnifica Humanitas (or Magnificent Humanity).
It is vital to mention that the Vatican stepped in by defending its conception of human dignity and the purpose of creation in the face of the AI revolution. Pope Leo XIV’s Magnifica Humanitas (Magnificent Humanity), the first encyclical of his pontificate, is quite ambitious: just watch its subtitle: “On Safeguarding the Human Person in the Time of Artificial Intelligence”.
More on Encyclical in: https://www.integrin.dk/2026/07/06/ai-and-higher-education-algorithmic-ethics-in-the-popes-encyclical/
By postulating that AI systems “merely imitate” human intelligence, the encyclical Magnifica Humanitas rejects the long-standing claim that imitating intelligence is the same as possessing it. Far from suggesting an absence of intelligence, that theory goes, imitation may be a facet of it. After all, humans also imitate intelligence—children learn from their parents, and professors master the teachings of earlier scholars. Moreover, if a machine can synthesize knowledge, converse cogently, analyze a medical scan, and ace the bar exam, then comes a question: in what sense is it not intelligent?
Source and citation from: https://www.foreignaffairs.com/reviews/god-machine-sebastian-mallaby
Addressing AI risks
Recently, the MIT AI Risk Initiative released another report that revealed how 272 experts assessed the severity of AI risks across various sectors, as well as how to mitigate them. Below are extracts from the AI-2026 report: more than 4 in 5 organizations are confident they can prevent unauthorized data access; yet among these organizations, AI-related unauthorized access incidents still affect 62% to 72% of respondents. Data security and privacy are the top concerns for both generative AI and AI agents. Securing data used for AI training is also the top future investment priority, cited by 4 in 5 organizations. Up to 1 in 5 organizations do not know whether employees are using unsanctioned AI tools.
For generative AI, that figure has nearly tripled since 2025. Nearly 9 in 10 organizations delayed both agentic and generative AI deployments by an average of almost six months; about 36 percent of enterprise data is already AI-generated, and respondents expect that share to reach over 42 percent within a year.
Nearly half of employees rely on AI agents weekly or daily; yet nearly 9 in 10 organizations experienced at least one agent-related security incident in the past year.
More in: The State of AI 2026: Scaling Trust, Control and Readiness in the Agentic Era. In:
https://www.avepoint.com/shifthappens/reports/artificial-intelligence-report-2026?utm_source=google&utm_medium