Generative AI and the Talent Crunch: Competitive Implications in Labour Markets, recent trends and the Perils of Overzealousness

Generative Artificial Intelligence (AI) has swiftly evolved from a niche technological field into a key driver of economic and social transformation. The ability to disrupt established markets, create entirely new sectors, and redefine the boundaries of competition has attracted significant attention from competition authorities. In line with other national competition authorities, the Portuguese Competition Authority (PCA) has been closely monitoring the competitive dynamics and risks associated with AI, paying particular attention to issues such as talent, market concentration, and the phenomenon of so-called “killer acquisitions”.

In July 2025, the PCA published the third in a series of short papers examining competition issues in the AI sector, with a particular emphasis on labour markets. This paper explores the importance of talent and know-how in AI development, the current shortage of specialised professionals, and the competitive implications of the various strategies adopted by AI developers to attract and retain talent. It also analyses the regulatory framework governing employment clauses and inter-firm agreements that may restrict labour mobility, as well as the increasing scrutiny of strategic acquisitions, such as “acquihires”. The main takeaways are:

(i) Access to Talent and Labour Mobility: Know-how, encompassing both explicit and tacit knowledge, is a fundamental input in the development of generative AI systems. The ability of AI developers to attract and retain skilled employees is a key determinant of competitive advantage, as innovation in this sector is often driven by the expertise and experimentation of individual developers and teams.

Labour mobility is identified as a primary channel for the dissemination of knowledge and innovation. The paper highlights that former employees of established tech firms have founded many leading AI companies, and that high rates of labour mobility, particularly in regions such as California, have contributed to the sector’s dynamism. Empirical evidence supports the view that labour mobility enhances productivity and innovation across the digital economy.

(ii) Talent Shortage and Competitive Dynamics: the generative AI sector is currently experiencing a significant shortage of highly specialised talent. This shortage is attributed to surging demand for AI expertise, the rapid pace of technological advancement, and the concentration of know-how within leading development teams. Large incumbents can offer substantial compensation packages, further exacerbating the challenges faced by smaller firms and academia in attracting and retaining talent.

In response, AI developers have adopted various strategies to secure access to talent, including direct recruitment from competitors and academia, as well as more complex approaches such as acquihires.

(iii) Acquihires and Merger Control: Acquihires—transactions in which a company hires a significant portion or all of another company’s workforce, often accompanied by licensing of intellectual property—have become increasingly prevalent in the AI sector. These transactions blur the line between traditional hiring and mergers, raising questions regarding their treatment under competition law.

Recent high-profile examples, such as the Microsoft/Inflection transaction, have prompted competition authorities to assess whether such deals constitute a “concentration” under the EU Merger Regulation (EUMR) and national competition laws. The transfer of key personnel and know-how may amount to a de facto transfer of control, even in the absence of a formal acquisition of assets. However, many acquihire transactions fall below traditional notification thresholds, which presents challenges for effective merger control. The paper notes that alternative notification criteria, such as market share or transaction value thresholds and call-in, are increasingly relevant in addressing these gaps.

(iv) Employment Clauses Restricting Labour Mobility: AI developers often include restrictive clauses in employment contracts, such as non-compete, invention assignment, intellectual property, and confidentiality provisions. While these clauses may incentivise investment in training and protect proprietary knowledge, they can also reduce labour mobility, limit knowledge spillovers, and potentially stifle innovation.

The prevalence of such clauses is particularly high in the generative AI and broader digital sectors. The PCA’s survey of firms in Portugal indicates the widespread use of non-compete and IP-related clauses, though non-solicitation and training repayment provisions are less common.

The legal frameworks governing these clauses vary across jurisdictions, typically setting limits on their scope, duration, and compensation. For instance, the Portuguese Labour Code restricts non-compete clauses to a maximum of two or three years, depending on the employee’s role and access to sensitive information, and requires proportional compensation.

From a competition law perspective, restrictive employment clauses generally fall outside the scope of Article 101 TFEU, as they are not agreements between undertakings. However, when used by dominant firms as part of an exclusionary strategy to limit competitors’ access to key talent, such clauses may attract scrutiny under abuse of dominant provisions (Article 102 TFEU).

(v) Inter-Firm Agreements Restricting Labour Mobility: agreements between firms that restrict labour mobility—such as no-poach and wage-fixing agreements—are considered per se violations of competition law under Article 9 of the Portuguese Competition Act and Article 101 Treaty on the Functioning of the European Union (TFEU). Such agreements are treated as restrictions by object, akin to buyers’ cartels, and are unlikely to qualify for exemption.

The PCA and other European competition authorities have issued guidance and taken enforcement action against such agreements, including fines in the IT and sports sectors. The European Commission has also brought its first cartel case involving a no-poach agreement in 2025.

In summary, the PCA’s analysis highlights the competition challenges arising from the generative AI sector’s dependence on specialised talent and proprietary know-how. These challenges are particularly acute in the context of talent shortages, restrictive employment practices, and strategic acquisitions. Labour mobility remains a critical driver of innovation and competition; in the PCA’s view, the increasing prevalence of restrictive covenants and acquihire strategies calls for vigilant regulatory oversight. Competition authorities are therefore adjusting their approaches to ensure that labour market practices do not unfairly hinder competition or stifle innovation in this rapidly evolving sector.

Recent developments show that national competition authorities (NCAs) are eager to engage proactively with the evolution of the AI market across all facets of competition law. While such regulatory activism is well-intentioned, it inevitably raises a number of fundamental questions. Foremost among these is whether authorities are intervening prematurely, acting on perceived or hypothetical risks rather than based on concrete evidence.

There is genuine concern that early or excessive intervention by NCAs could supress development of AI and hinder the sector’s growth trajectory. Many participants in this space are startups, often lacking the financial resources and investment capacity to scale independently. For these firms, partnerships or acquisitions frequently represent the most viable route to commercialisation. Overly aggressive scrutiny of so-called “killer acquisitions” risks undermining this dynamic, potentially hindering innovation and limiting the ability of promising AI start-ups to achieve their full market potential.

Merger control and regulatory reform

In the realm of merger control, NCAs in key jurisdictions are intensifying their scrutiny of “killer acquisitions”, leading to significant regulatory developments in the oversight of artificial intelligence from a competition standpoint. Such transactions—typically involving established players acquiring smaller, innovative targets—are viewed as a potential threat to innovation and market dynamism, as they may prevent new products from reaching the market or eliminate emerging competitors.

The primary concern is that these deals could hinder innovation and increase market concentration, ultimately to the detriment of consumers. However, traditional merger control regimes, which rely heavily on turnover thresholds, have often constrained the ability of authorities to intervene effectively. In response, there is a discernible trend towards reform, with authorities considering the introduction of deal value-based thresholds and making greater use of call-in powers for transactions falling below existing thresholds.

Within the European Union, the European Commission is actively exploring ways to strengthen its powers to review below-threshold transactions, particularly in the wake of the Illumina/Grail judgment, which limited the Commission’s broad interpretation of its referral powers. The European Commission is contemplating reforms such as the adoption of deal value-based thresholds and enhanced call-in powers under the EU Merger Regulation. Meanwhile, national competition authorities within the EU are expected to make increased use of their call-in powers, with the possibility of referring cases to the Commission.

In the United Kingdom, the Digital Markets, Competition and Consumers Act 2024 has introduced a new acquirer-focused jurisdictional threshold, empowering the Competition and Markets Authority (CMA) to review killer acquisitions across all sectors.

These reforms reflect a global shift towards more robust oversight of transactions that could harm innovation and competition, while also seeking to maintain legal certainty and avoid unduly discouraging M&A activity in innovative sectors.

The Path Forward: Striking the Right Balance

If, as appears likely, the PCA is echoing the broader European debate, it is clear that a variety of challenges exists throughout the AI lifecycle. The CMA, for its part, has adopted a notably proactive stance, intensifying scrutiny of Big Tech’s involvement in AI start-ups. The CMA’s recent “invitations to comment” on Microsoft’s investment in Mistral, the hiring of DeepMind’s co-founder, and Amazon’s investment in Anthropic are not mere formalities. Instead, they signal a willingness to intervene at an early stage—prior to the establishment of formal control—where there are concerns that such partnerships may confer de facto control or otherwise impede competition.

This recent surge in regulatory activity, while driven by legitimate concerns, prompts several important questions. Are authorities intervening too early, based on hypothetical risks rather than concrete evidence? The AI market is highly dynamic, with today’s leading firms potentially supplanted by new entrants tomorrow. Overzealous intervention could inadvertently stifle the very innovation it seeks to protect. Furthermore, could regulatory uncertainty deter investment and collaboration? If every partnership, minority investment, or key hire is subject to intense scrutiny, firms may become risk-averse, potentially hindering technological progress and ultimately harming consumer welfare. Finally, how can authorities ensure that intervention remains proportionate and evidence-based? The challenge lies in striking a delicate balance between vigilance and restraint—preserving competition without smothering the competitive process itself.

As the regulatory landscape continues to evolve, competition authorities must carefully calibrate their interventions, ensuring that they support, rather than hinder, the development and diffusion of generative AI technologies. The ultimate objective must be to foster an environment in which innovation can flourish, to the benefit of consumers and society as a whole.