28-07-2025

AI Safety refers to the practices and principles that help ensure that AI technologies are designed and used in ways that benefit humanity while minimizing any potential harm or negative outcomes.
Building safe artificial intelligence (AI) systems is a critical consideration for businesses and society, given the increasing prevalence and impact of AI. AI safety ensures that systems are used as responsibly as possible and that the future of this technology develops in alignment with human values.
Developing and maintaining safe AI involves identifying potential risks (such as bias, data security, and vulnerability to external threats) and creating processes to prevent and mitigate those risks.

Why is AI safety important?
As AI systems become more sophisticated, they are increasingly integrated into people’s lives and into critical real-world areas such as infrastructure, finance, and national security. These technologies can have both positive and negative repercussions on the organizations that use them and on society as a whole.
Experts are also concerned that some advanced AI systems may eventually become as intelligent as—or even more intelligent than—humans. Artificial General Intelligence (AGI) refers to potential AI systems that can understand, learn, and perform cognitive tasks in the same way humans do. Artificial Superintelligence (ASI) refers to hypothetical AI systems with intellectual capabilities and cognitive functions more advanced than any human being. The development of AGI and ASI raises concerns that these systems could be dangerous if not aligned with human values or not subject to human oversight. If granted too much autonomy, critics argue that such systems could pose an existential threat to humanity.
From a business perspective, safe AI helps build consumer trust, protects against legal liabilities, and avoids flawed decision-making. Organizations that take steps to ensure AI use is aligned with their values can prevent negative consequences for themselves and their clients.

Types of AI Risks
AI risks can be classified into several categories, each requiring different AI safety measures and risk management efforts:

• Bias and fairness

• Privacy

• Loss of control

• Existential risks

• Malicious misuse

• Cybersecurity

Bias and Fairness
AI systems can perpetuate or amplify social biases. Algorithmic bias occurs when AI is trained on incomplete or misleading data. This can result in unfair decision-making. For instance, an AI tool trained on discriminatory data might be less likely to approve mortgages for applicants from certain backgrounds, or more likely to recommend hiring a man instead of a woman.

Privacy
AI systems can access, expose, or misuse personal data, raising privacy concerns. If sensitive data is compromised, AI developers or users may be held accountable.

Loss of Control
Outcomes from advanced AI systems—particularly those designed to act as autonomous agents—can be unpredictable. Their actions may also be harmful. If capable of making independent decisions, stopping them may prove difficult. Without human oversight, it could be impossible to intervene or shut down a misbehaving AI system.

Existential Risks
AGI, ASI, and other highly advanced AI systems could act in ways that endanger humanity or disrupt global systems if mismanaged.
AI could also be misused for large-scale social manipulation or cyber warfare. In 2023, the Center for AI Safety (CAIS), a nonprofit organization, published a one-sentence statement endorsed by several AI researchers and leaders: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

Malicious Misuse
While unintended consequences and errors are a source of AI risk, bad actors may also intentionally use the technology to cause harm. AI can be weaponized for cyberattacks, disinformation campaigns, illegal surveillance, or even physical harm. These threats exist both at an individual and societal level.

Cybersecurity
AI systems may be vulnerable to security threats. They face risks of adversarial attacks, where malicious actors manipulate data inputs to mislead models and produce incorrect outputs.
For example, AI jailbreaks occur when hackers exploit vulnerabilities through prompt injections or other techniques to force systems into restricted actions. Data poisoning occurs when compromised training data skews AI behavior. Unauthorized access and other security flaws may lead to the misuse of AI systems and their data.

In an upcoming bulletin, we will continue to explore this topic, which is regarded as a critical, complex, and dynamic area that is constantly evolving due to ethical challenges, the pace of technological advancements, and the need to protect privacy and data integrity.
AI safety is not merely a technical issue but a fundamental pillar for the ethical, responsible, and beneficial development of artificial intelligence in the future. It is essential to address it proactively to minimize risks and maximize the positive potential of this technology.

If you have any questions regarding this matter, please do not hesitate to contact us at tel.: 7078 8001 or by email at it@jebsen.com.ar.

Sincerely,
Laura Borroni