The EU AI Act’s August 2 enforcement window is 105 days away, and this week we got specifics that matter. An AI hiring bias audit is no longer a theoretical compliance exercise — regulators confirmed the exact audit scope, documentation requirements, and the lead-time problem with certified auditors. Add Anthropic’s new flagship model and OpenAI’s record $122 billion raise to the mix, and you have four things worth understanding before the week ends.

EU Mandates AI Hiring Bias Audits — €15M Penalty, 105 Days Left

From August 2, 2026, any AI system used in employment decisions falls under the EU AI Act’s high-risk category. That means annual third-party AI hiring bias audits, full technical documentation, human oversight mechanisms, and transparency disclosures to candidates. The penalty for non-compliance: €15 million or 3% of global annual turnover, whichever is higher. (Source: AI2Work)

The scope is broader than most HR leaders realize. The rule covers any AI tool used to evaluate EU-based candidates — regardless of where your company is headquartered. Your ATS’s resume-ranking algorithm, your AI interview scoring tool, your job ad targeting system: all of it qualifies as high-risk under Annex III of the Act. If you’re a US or India-based company with remote roles open to EU applicants, this applies to you.

An AI hiring bias audit isn’t a one-page checkbox. Auditors need your model’s training data composition, outcome data disaggregated by protected characteristics (gender, age, ethnicity), and evidence of continuous monitoring. Most companies haven’t built the data pipelines for this. And “we use a vendor’s tool” is not a defense — the obligation falls on the deployer, not the developer. (Source: Raconteur)

There’s also a booking problem. Certified third-party auditors qualified under the EU’s conformity assessment framework are limited in number, and they’re filling up. Companies targeting August 2 compliance need an audit engagement letter signed now, not in July.

If you’re running a 100-person company that started using AI-based resume screening in the last year, here’s what you need before the deadline: a written risk assessment, documentation of your training data sources, a human review process for rejected candidates, and a signed audit engagement. Companies already building AI into their HR recruitment workflows need to layer these documentation requirements on top of their existing processes, not treat them as a separate project.

What to do: Ask your ATS or AI hiring vendor directly: are you classified as a high-risk system under Annex III of the EU AI Act? If they don’t know, that’s a red flag. Start with that conversation this week.

EU AI Act Article 12: Every AI Agent Action Must Now Be Logged

Same August 2 deadline, different obligation. Under Article 12 of the EU AI Act, high-risk AI agents must automatically log every action they take, with enough detail to trace their decision-making. Logs must be retained for at least six months, be structured and searchable, and attributable to a specific agent and action on demand. (Source: Security Boulevard / FireTail)

This week’s technical guidance clarified what “automatic” means in practice: logs generated without operator intervention at the moment events occur. Not batch exports. Not audit trails compiled on request. Real-time, always-on records of what your agent did and why.

For companies already working through their AI hiring bias audit obligations, the agent logging requirement is a separate but related compliance track. If you’re deploying AI agents for HR workflows — onboarding bots, benefits enrollment agents, performance review assistants — ask your vendor whether their system produces Article 12-compliant logs today. Most don’t yet. There’s no finalized technical standard (draft frameworks prEN 18229-1 and ISO/IEC DIS 24970 are still under review), which means you’ll need to define your own logging architecture in the interim.

Anthropic Ships Claude Opus 4.7: Better Coding, Vision, and a Differentiated Safety Model

Anthropic released Claude Opus 4.7 on April 16. Better agentic coding, sharper multidisciplinary reasoning, scaled tool use, and one headline feature: high-resolution image support up to 2576px / 3.75MP — the first Claude model to offer it. Pricing is unchanged from Opus 4.6. (Source: CNBC)

The safety angle matters for enterprise buyers. Anthropic “differentially reduced” Opus 4.7’s cybersecurity capabilities during training — a direct response to dual-use risk concerns. Security teams who need full cyber capabilities for legitimate research can apply through a formal verification program. For HR and ops teams building agentic workflows, that’s not a blocker. But it signals a direction: AI labs are moving toward differentiated capability tiers based on verified use case, not a single model for everything.

Worth noting: Anthropic’s unreleased model, codenamed Mythos, reportedly outperforms Opus 4.7 on key benchmarks. Anthropic acknowledged this in the launch materials. So what you’re getting today is a strong production model — just not their ceiling. For teams evaluating top AI tools for HR, Opus 4.7 is a practical choice now, with better options likely coming later this year.

OpenAI Closes $122B at $852B Valuation — Enterprise HR Tools Next

OpenAI completed its $122 billion funding round on March 31, locking in an $852 billion valuation. Amazon anchored at $50 billion, Nvidia at $30 billion, SoftBank at $30 billion. OpenAI also opened part of the round to individual investors through bank channels for the first time — a clear signal that a public offering is close. The company is reporting $2 billion in monthly revenue and over 900 million weekly active users, with enterprise making up more than 40% of that revenue. (Source: TechCrunch, Bloomberg)

For HR leaders, the practical implication: the AI hiring bias audit conversation and workforce analytics investment happening inside OpenAI’s enterprise roadmap just got a very large capital injection to accelerate it. The products coming out of this round won’t stay optional for long. They’ll be table stakes, and the EU compliance clock is already running.

If navigating AI hiring bias audit requirements has you rethinking how your team evaluates and deploys AI tools, Asanify’s HR platform is built for AI in human resource management with compliance workflows in mind. For teams starting fresh on tool evaluation, our guide to AI applicant tracking systems covers what documentation you’ll need under the new rules.

Frequently Asked Questions About AI Hiring Bias Audits

What is an AI hiring bias audit under the EU AI Act?

An AI hiring bias audit is an annual review by a certified third party confirming that your AI recruitment tools don’t produce discriminatory outcomes. Required for any system that screens resumes, scores interviews, or targets job ads at EU-based candidates, the audit covers training data composition, outcome data disaggregated by protected characteristics, and evidence of continuous monitoring.

Does the EU AI Act apply to companies headquartered outside Europe?

Yes. The Act covers any organization whose AI systems evaluate EU-based candidates or employees, regardless of where the company is headquartered. A startup based in India or the US that posts remote roles open to EU applicants falls within scope if their AI tools make decisions affecting EU individuals.

What’s the penalty for missing the August 2, 2026 EU AI Act deadline?

Non-compliance with high-risk system obligations carries fines of up to €15 million or 3% of global annual turnover, whichever is higher. For prohibited AI practices — such as real-time biometric surveillance in public spaces — the ceiling rises to €35 million or 7% of turnover. Enforcement authority sits with national AI supervisory authorities in each EU member state.

Not to be considered as tax, legal, financial or HR advice. Regulations change over time so please consult a lawyer, accountant  or Labour Law  expert for specific guidance.