AWS Rewrites Three Certifications Around GenAI and Agentic AI

On September 1, 2026, AWS Training and Certification published an update to three certification exams: AWS Certified Machine Learning Engineer Associate (MLA-C02), AWS Certified Solutions Architect Professional (SAP-C03) and AWS Certified Developer Associate (DVA-C03). The stated reason is consistent across all three: cloud roles have changed. Machine learning engineers no longer only build traditional models, developers no longer only write application code, and architects are expected to design generative AI and agentic architectures into the stack. In effect, three exams were rewritten in the same month by the same force, moving generative AI, agentic AI and AI-assisted development into the core blueprint.

1. Three Exams, One Direction

The shared direction is worth seeing before the details. MLA-C02 redefines the machine learning engineer as someone who covers traditional ML and generative AI together. SAP-C03 adds generative AI and agentic AI, resilience engineering and post-quantum cryptography to the architect's blueprint. DVA-C03 makes AI-assisted development a core competency rather than an optional add-on, in AWS's own words. So if your credential dates from 2024 or 2025, the question is not merely whether to recertify; it is which capability the new exam actually tests.

# Weight your study time, do not spread it evenly. DVA-C03 keeps four
# domains and their percentages are unchanged from DVA-C02, so the new
# work is inside the existing domains rather than a new domain.

DOMAINS = {
    "Development with AWS Services": 30,   # + AI-assisted code gen and review
    "Security": 26,                        # + new Task 2.3: AI security
    "Testing and Deployment": 22,          # + AI-assisted test automation, CI/CD
    "Troubleshooting and Optimization": 22 # + AI-assisted error and perf analysis
}

def plan(weeks_available, per_week):
    total = weeks_available * per_week
    return {d: round(total * w / 100, 1) for d, w in DOMAINS.items()}

print(plan(8, 6))   # {'Development with AWS Services': 14.4, 'Security': 12.5, ...}

Three exams, one direction

2. MLA-C02: The ML Engineer Becomes GenAI Plus Agentic

Start with MLA-C02. Beta registration opened in English on September 1, 2026. The last day to take the current version, MLA-C01, in English is September 28, 2026, and beta delivery begins on September 29. General availability starts on January 14, 2027 in all languages, at which point MLA-C01 retires in every language. The domain structure is unchanged and no new domains were added, but the content expands along practice: generative AI implementation including fine-tuning foundation models and building retrieval-augmented generation architectures, agentic AI including orchestrating agents and complex workflows, foundation models and LLMs including selecting, customising and operationalising them, expanded coverage of Amazon Bedrock, and updated responsible AI practices. The beta is 170 minutes, 85 questions, USD 75, English only, delivered through Pearson VUE.

# SAP-C03 adds 11 skills for "GenAI and agentic AI". One of them is agent
# architectures with Amazon Bedrock AgentCore. Sketch the smallest agent
# you can explain out loud, because the exam asks you to reason about it.

import boto3

client = boto3.client("bedrock-agentcore")

agent = client.create_agent(
    agentName="invoice-review-agent",
    foundationModel="amazon.nova-pro-v1:0",
    instruction="Extract line items, flag mismatches, never approve payment.",
    actionGroups=[{                      # what the agent may actually do
        "actionGroupName": "read-invoice",
        "apiSchema": {"payload": "s3://schemas/read-invoice.json"},
    }],
)

# Exam themes to be able to argue: least privilege per action group,
# guardrails for content filtering, and a human-oversight workflow for the
# one action you will not let a model take unattended.

3. SAP-C03: Architects Get Agents and Post-Quantum Crypto

Now SAP-C03. Registration opens on October 27, 2026; the last day for SAP-C02 is November 16, 2026; SAP-C03 goes GA on November 17. The exam is 180 minutes, 75 questions and USD 300. The heaviest addition is generative AI and agentic AI with 11 new skills, including designing generative AI integrations with Amazon Bedrock, agent architectures with Amazon Bedrock AgentCore, RAG architectures, content filtering with Bedrock Guardrails, and human-oversight workflows for AI operations. The rest covers resilience engineering with AWS Fault Injection Service, Resilience Hub, Application Recovery Controller and Systems Manager runbooks; cloud-native patterns including durable workflows with Step Functions, service mesh with VPC Lattice and ECS Service Connect, container image security, multi-tenant and real-time data architectures; DevSecOps and observability; data and analytics with Lake Formation and Apache Iceberg plus AWS Clean Rooms; and post-quantum cryptography using AWS KMS with ML-DSA and ML-KEM.

# The MLA-C02 beta adds retrieval-augmented generation as an
# implementation skill. Keep the pattern explicit: retrieve, augment,
# generate, and never let the retriever write.

def rag_answer(question, retriever, model, k=5):
    chunks = retriever.search(question, top_k=k)      # read-only step
    context = "\n\n".join(f"[{i}] {c.text}" for i, c in enumerate(chunks))
    prompt = (
        "Answer using ONLY the numbered context. Cite [n] for every claim.\n\n"
        f"{context}\n\nQuestion: {question}"
    )
    answer = model.generate(prompt, max_tokens=400)
    return {"answer": answer, "sources": [c.id for c in chunks]}
    # Grader-facing truth: RAG is a data problem first; cite or refuse.

Bedrock, AgentCore and SageMaker AI enter the blueprint

4. DVA-C03: Developers Get AI-Assisted Development and AI Security

DVA-C03 is the most directly relevant to developers. Registration opens on October 27, 2026, the last day for DVA-C02 is November 30, 2026, and DVA-C03 goes GA on December 1. The exam is 130 minutes, 65 questions with 50 scored and 15 unscored, costs USD 150, and requires 720 out of 1000 to pass. The four domains and their weightings are unchanged at 30 percent development, 26 percent security, 22 percent testing and deployment, and 22 percent troubleshooting and optimisation. Inside security, however, there is a new Task 2.3 on AI security with five skills: access management for AI services, data privacy controls including VPC endpoints and ensuring inputs and outputs are not used to train models, content filtering and prompt-injection protection, AI agent interaction security covering tool-use authorisation, session isolation and human-in-the-loop approval, and protecting sensitive content in monitoring logs. Container management with ECR, ECS, EKS and Fargate is added, along with new in-scope services: Amazon Bedrock, Amazon Bedrock AgentCore, Amazon Q, Kiro, Amazon Data Firehose and AWS PrivateLink. Note what is explicitly out of scope: prompt engineering, RAG design, AI model selection, Amazon SageMaker AI and AI governance framework design.

# DVA-C03's new Task 2.3 is AI security, and one of its five skills is
# prompt-injection protection. Treat every model output that names a tool
# as untrusted input, because that is what it is.

ALLOWED_TOOLS = {"read_ticket", "search_kb"}

def guard_tool_call(call, allow=ALLOWED_TOOLS):
    if call["tool"] not in allow:
        return {"decision": "deny", "reason": "tool not in allowlist"}
    if looks_like_injection(call.get("args", {})):
        return {"decision": "deny", "reason": "injection pattern in arguments"}
    return {"decision": "allow", "scope": call["tool"]}

def looks_like_injection(args):
    blob = " ".join(str(v) for v in args.values()).lower()
    return any(s in blob for s in ("ignore previous", "system:", "exfiltrate"))

# A session that cannot be reused for a different user is the second half
# of the same skill: session isolation plus tool-use authorisation.

5. Key Dates at a Glance

The dates deserve their own list, because the old exams disappear on a schedule and a missed window resets your plan. MLA-C02 beta registration opened on September 1, 2026; MLA-C01's last English sitting is September 28; beta delivery begins September 29; MLA-C02 reaches GA and MLA-C01 retires in all languages on January 14, 2027. Registration for SAP-C03 and DVA-C03 both open on October 27, 2026. The final dates for the old versions are November 16 for SAP-C02 and November 30 for DVA-C02. The new versions go GA on November 17 for SAP-C03 and December 1 for DVA-C03. Code sample 5 turns that schedule into a calendar file you can share with a team.

# The dates matter more than the topics, because the old exams disappear
# on a schedule. Generate a local calendar so nobody on the team misses a
# registration window or a last-chance sitting.

EVENTS = {
    "MLA-C02 beta registration opens":       "2026-09-01",
    "MLA-C01 last day (English)":            "2026-09-28",
    "MLA-C02 beta delivery begins":          "2026-09-29",
    "MLA-C02 GA, MLA-C01 retired":           "2027-01-14",
    "SAP-C03 + DVA-C03 registration opens":  "2026-10-27",
    "SAP-C02 last day":                      "2026-11-16",
    "SAP-C03 GA":                            "2026-11-17",
    "DVA-C02 last day":                      "2026-11-30",
    "DVA-C03 GA":                            "2026-12-01",
}

def ics(events):
    lines = ["BEGIN:VCALENDAR", "VERSION:2.0", "PRODID:-//certs//EN"]
    for title, day in events.items():
        d = day.replace("-", "")
        lines += ["BEGIN:VEVENT", f"DTSTART;VALUE=DATE:{d}",
                  f"SUMMARY:{title}", "END:VEVENT"]
    return "\n".join(lines + ["END:VCALENDAR"])

print(ics(EVENTS)[:60])

DVA-C03 adds an AI security task, including prompt-injection protection

6. Should You Take the Old Exam or the New One?

Finally, how to choose. AWS gives a clear rule: take the current version if you are already prepared and want the credential now, since your credential remains active through its original expiration date, and wait for the updated version if you want the certification to validate generative AI, agentic AI and modern security practices. For an individual, a practical test is whether an interviewer would probe the new material. The AI security and AI-assisted development skills in DVA-C03, and the agent architecture and post-quantum content in SAP-C03, all describe problems that appear in production today. MLA-C02 suits people moving from data science toward LLMOps. For a team, first count how many people can sit the old exam before its final day, then decide whether to sprint on the old version or switch wholesale to the new one.