Amnesty International Index in Stanford: 5 critical visions to reshape the technical strategy of institutions

The Stanford Institute of AI focused on human (HiIt issued the AI ​​2025 index report, providing data -based analysis for the global development of Amnesty International. Hai has developed a report on artificial intelligence over the past few years, with it The first standard He is coming in 2022. It goes without saying, a lot has changed.

The 2025 report is loaded with statistics. Among some of the higher results:

• The United States produced 40 prominent Amnesty International models in 2024, before China (15) and Europe (3).
• The training account for artificial intelligence models doubles almost every five months, and the sizes of data sets every eight months.
• The costs of the conclusion of the artificial intelligence model decreased significantly-a decrease in 280 times from 2022 to 2024.
• Global Private Ai investing $ 252.3 billion in 2024, an increase of 26 %.
• 78 % of organizations are reports using artificial intelligence (an increase of 55 % in 2023).

For information technology leaders for institutions who sprinkle the strategy of artificial intelligence, the report provides decisive visions of typical performance, investment trends, implementation challenges and competitive dynamics to reshape the technological scene.
Here are five main meals for the leaders of the Information Technology Corporation from the AI ​​Index.

1.

Perhaps the most surprising result is the extent of high -quality artificial intelligence at reasonable and accessible prices. The cost barrier that restricts advanced artificial intelligence on technology giants collapses. The result contrasts starkly with what 2024 Stanford Report He found.

“I am surprised how much artificial intelligence models have become cheaper, more open and accessible over the past year,” told Nestor Maslej, director of the Hai Venturebeat Research Index. “Although the training costs are still high, we now see a world where the cost of developing high quality-despite the lack of limits-decreases.

The report specifies this transformation significantly: The cost of inferring the artificial intelligence model that leads to GPT-3.5 levels of $ 20.00 per million symbols in November 2022 decreased to only $ 0.07 per million symbols by October 2024-declining 280 times in 18 months.

Equally important, the convergence of performance between closed and open models. The gap between the upper closed models (such as GPT-4) and the leading open models (such as Llama) narrowed from 8.0 % in January 2024 to only 1.7 % by February 2025.

IT Commander Work componentRe -evaluating your artificial intelligence strategy. Organizations previously priced in advanced artificial intelligence capabilities now have viable options through open -weight models or commercial application programming facades much cheaper.

2. The gap between the adoption of artificial intelligence and the achievement of value is still large

While the report shows that 78 % of institutions now use artificial intelligence in at least one job function (an increase of 55 % in 2023), the effect of real business is behind adoption.

When asked about a widespread meaningful investment return, Maslege admitted: “We have limited data about what separates organizations that achieve huge returns for timing with artificial intelligence from those not.

The report indicates that most of the institutions that use the Amnesty International report reports modest financial improvements. For example, 47 % of companies that use obstetric intelligence in strategy and increase corporate financing reports, but usually at levels less than 5 %.

IT Commander Work componentFocusing on measurable cases with the capabilities of the return on clear investment instead of widespread implementation. Consider the development of governance and measurement frameworks is stronger than artificial intelligence to better create value creation.

3. Specific business jobs show stronger financial returns than artificial intelligence

The report provides beloved visions in which business function sees the most important financial influence than the implementation of artificial intelligence.

“On the cost side, Amnesty International appears to benefit from the functions of the supply chain and service operations more than others,” Maslege pointed out. “On the aspect of revenue, you see the strategy, corporate financing, and supply chain functions of the biggest gains.”

Specifically, 61 % of the institutions that use artificial intelligence in the supply chain and the report of cost savings, while 70 % use it in strategy and financing corporate reports. Service and marketing/sales also shows strong potential to create value.

IT Commander Work component: Determining the priorities of artificial intelligence investments in jobs that show the most fundamental financial returns in the report. Improvement of supply chain, service and strategic planning processes as highly capabilities to spread initial or expanded artificial intelligence.

4. Amnesty International shows strong potential to achieve equality of workforce

One of the most interesting results related to the effect of artificial intelligence on the workforce productivity through skills levels. Multiple studies mentioned in the report show that artificial intelligence tools are not commensurate with low skills.

In customer support contexts, low workers have achieved 34 % skill of productivity with the help of artificial intelligence, while high skills workers have seen the minimum improvement. Similar patterns appeared in consulting (43 % compared to 16.5 % gains) and software engineering (21-40 % compared to 7-16 % gains).

“In general, these studies indicate that artificial intelligence has strong positive effects on productivity and tends to take advantage of low -skilled workers than that higher skill, but not always,” Masslig explained.

IT Commander Work component: Consider spreading artificial intelligence as a strategy for the development of workforce. Help with artificial intelligence can help settle the stadium between beginners and senior employees, and may treat skills gaps while improving the team’s performance in general.

5. The implementation of AI is still ambitious, not a fact

Despite the increasing awareness of the risks of artificial intelligence, the report reveals a large gap between risk recognition and mitigation. While 66 % of organizations consider cybersecurity a risk associated with AI, only 55 % actively relieve them. There are gaps similar to organizational compliance (63 % compared to 38 %) and violation of intellectual property (57 % compared to 38 %).

These results come against a background to increase artificial intelligence incidents, which increased by 56.4 % to 233 cases reported in 2024. Organizations face real consequences for their failure to implement responsible artificial intelligence practices.

IT Commander Work component: Do not delay the implementation of strong artificial intelligence governance. While technical capabilities are progressing quickly, the report indicates that most organizations still lack effective strategies to relieve risk. The development of these frameworks can now be a competitive advantage rather than the burden of compliance.

We look forward

The STANFORD AI report provides a picture of artificial intelligence technology at a more easy and capable speed, while organizations are still struggling to benefit from their entire potential.

For IT leaders, the strategic necessity is clear: focusing on the targeted applications with a measurable investment return, confirms responsible governance and benefit from artificial intelligence to enhance the capabilities of the workforce.

“This shift indicates more access, and I think a wave of broader artificial intelligence may be on the horizon,” Maslege said.