Meta plans to replace its human content security employees with AI...
According to internal documents obtained by NPR, Meta is phasing out its human-led product risk assessments in favor of AI-powered automation. This move represents a significant shift in how the tech giant, which owns Facebook, Instagram, and WhatsApp, evaluates the potential risks associated with its product updates and features.
Automation of Product Reviews
For more than a decade, Meta has relied on human reviewers to conduct "privacy and integrity reviews" internally. These reviews were designed to ensure that new features did not compromise user privacy, promote harmful content, or put young users at risk. However, NPR reports that up to 90% of these reviews will soon be automated using AI.
This automation will extend to changes in algorithms, new sharing options, as well as features related to youth safety and AI ethics. Meta plans to utilize AI tools in evaluating the potential risks posed by their products, with minimal human intervention.
Speeding up Product Development
Meta believes that this shift to AI will accelerate product development. Developers are expected to receive rapid AI-generated feedback based on questionnaires they submit about new products. The AI systems will identify potential risks and outline necessary steps to mitigate them, which the teams must address before launching the products.
The company states that human reviewers will still be involved in reviewing complex or novel cases, with only low-risk decisions being automated. This approach allows human reviewers to focus on more critical or ambiguous content moderation issues.
Expanding Use of AI
Meta's adoption of AI extends beyond product risk assessments. CEO Mark Zuckerberg recently announced plans for AI to play a more significant role in the company's operations. Within the next 12 to 18 months, a substantial portion of Meta's AI code, including its Llama models, will be generated by AI itself. Zuckerberg highlighted the AI agents' capabilities in conducting tests, identifying bugs, and producing high-quality code.
Meta's embrace of AI reflects a broader trend in the tech industry. Google's Sundar Pichai disclosed that AI now writes 30% of the company's code, while OpenAI's Sam Altman mentioned that half of all code in some companies is AI-generated. Anthropic CEO Dario Amodei predicted that by the end of 2025, almost all code will be generated by AI.
The shift towards AI-driven decision-making raises concerns about the potential loss of human oversight. While Meta claims that AI-made decisions are being audited, some insiders suggest that most risk decisions are already being entrusted to algorithms globally.
A current Meta employee emphasized the goal of empowering product teams and streamlining decision-making processes. However, they cautioned that eliminating human perspectives in risk assessment could have severe consequences.
