diff --git a/ast/builtins.go b/ast/builtins.go
index e3f47ac3d9..1ab7dc0a93 100644
--- a/ast/builtins.go
+++ b/ast/builtins.go
@@ -50,6 +50,14 @@ var DefaultBuiltins = [...]*Builtin{
 	Abs,
 	Rem,
 
+	// Bitwise Arithmetic
+	BitsOr,
+	BitsAnd,
+	BitsNegate,
+	BitsXOr,
+	BitsShiftLeft,
+	BitsShiftRight,
+
 	// Binary
 	And,
 	Or,
@@ -383,10 +391,69 @@ var Rem = &Builtin{
 }
 
 /**
- * Binary
+ * Bitwise
  */
 
-// TODO(tsandall): update binary operators to support integers.
+// BitsOr returns the bitwise "or" of two integers.
+var BitsOr = &Builtin{
+	Name: "bits.or",
+	Decl: types.NewFunction(
+		types.Args(types.N, types.N),
+		types.N,
+	),
+}
+
+// BitsAnd returns the bitwise "and" of two integers.
+var BitsAnd = &Builtin{
+	Name: "bits.and",
+	Decl: types.NewFunction(
+		types.Args(types.N, types.N),
+		types.N,
+	),
+}
+
+// BitsNegate returns the bitwise "negation" of an integer (i.e. flips each
+// bit).
+var BitsNegate = &Builtin{
+	Name: "bits.negate",
+	Decl: types.NewFunction(
+		types.Args(types.N),
+		types.N,
+	),
+}
+
+// BitsXOr returns the bitwise "exclusive-or" of two integers.
+var BitsXOr = &Builtin{
+	Name: "bits.xor",
+	Decl: types.NewFunction(
+		types.Args(types.N, types.N),
+		types.N,
+	),
+}
+
+// BitsShiftLeft returns a new integer with its bits shifted some value to the
+// left.
+var BitsShiftLeft = &Builtin{
+	Name: "bits.lsh",
+	Decl: types.NewFunction(
+		types.Args(types.N, types.N),
+		types.N,
+	),
+}
+
+// BitsShiftRight returns a new integer with its bits shifted some value to the
+// right.
+var BitsShiftRight = &Builtin{
+	Name: "bits.rsh",
+	Decl: types.NewFunction(
+		types.Args(types.N, types.N),
+		types.N,
+	),
+}
+
+/**
+ * Sets
+ */
 
 // And performs an intersection operation on sets.
 var And = &Builtin{
diff --git a/docs/content/policy-reference.md b/docs/content/policy-reference.md
index 21a9092202..a3cf6ad954 100644
--- a/docs/content/policy-reference.md
+++ b/docs/content/policy-reference.md
@@ -161,6 +161,17 @@ The following table shows examples of how ``glob.match`` works:
 | ``output := glob.match("{cat,bat,[fr]at}", [], "rat")`` | ``true`` | A glob with pattern-alternatives matchers. |
 | ``output := glob.match("{cat,bat,[fr]at}", [], "at")`` | ``false`` | A glob with pattern-alternatives matchers. |
 
+### Bitwise
+
+| Built-in | Description |
+| --- | --- |
+| <span class="opa-keep-it-together">``z := bits.or(x, y)``</span>  | ``z`` is the bitwise or of integers ``x`` and ``y`` |
+| <span class="opa-keep-it-together">``z := bits.and(x, y)``</span> | ``z`` is the bitwise and of integers ``x`` and ``y`` |
+| <span class="opa-keep-it-together">``z := bits.negate(x)``</span> | ``z`` is the bitwise negation (flip) of integer ``x`` |
+| <span class="opa-keep-it-together">``z := bits.xor(x, y)``</span> | ``z`` is the bitwise exclusive-or of integers ``x`` and ``y`` |
+| <span class="opa-keep-it-together">``z := bits.lsh(x, s)``</span> | ``z`` is the bitshift of integer ``x`` by ``s`` bits to the left |
+| <span class="opa-keep-it-together">``z := bits.rsh(x, s)``</span> | ``z`` is the bitshift of integer ``x`` by ``s`` bits to the right |
+
 ### Conversions
 
 | Built-in | Description |
diff --git a/topdown/bits.go b/topdown/bits.go
new file mode 100644
index 0000000000..7a63c0df1e
--- /dev/null
+++ b/topdown/bits.go
@@ -0,0 +1,88 @@
+// Copyright 2020 The OPA Authors.  All rights reserved.
+// Use of this source code is governed by an Apache2
+// license that can be found in the LICENSE file.
+
+package topdown
+
+import (
+	"math/big"
+
+	"github.com/open-policy-agent/opa/ast"
+	"github.com/open-policy-agent/opa/topdown/builtins"
+)
+
+type bitsArity1 func(a *big.Int) (*big.Int, error)
+type bitsArity2 func(a, b *big.Int) (*big.Int, error)
+
+func bitsOr(a, b *big.Int) (*big.Int, error) {
+	return new(big.Int).Or(a, b), nil
+}
+
+func bitsAnd(a, b *big.Int) (*big.Int, error) {
+	return new(big.Int).And(a, b), nil
+}
+
+func bitsNegate(a *big.Int) (*big.Int, error) {
+	return new(big.Int).Not(a), nil
+}
+
+func bitsXOr(a, b *big.Int) (*big.Int, error) {
+	return new(big.Int).Xor(a, b), nil
+}
+
+func bitsShiftLeft(a, b *big.Int) (*big.Int, error) {
+	if b.Sign() == -1 {
+		return nil, builtins.NewOperandErr(2, "must be an unsigned integer number but got a negative integer")
+	}
+	shift := uint(b.Uint64())
+	return new(big.Int).Lsh(a, shift), nil
+}
+
+func bitsShiftRight(a, b *big.Int) (*big.Int, error) {
+	if b.Sign() == -1 {
+		return nil, builtins.NewOperandErr(2, "must be an unsigned integer number but got a negative integer")
+	}
+	shift := uint(b.Uint64())
+	return new(big.Int).Rsh(a, shift), nil
+}
+
+func builtinBitsArity1(fn bitsArity1) BuiltinFunc {
+	return func(_ BuiltinContext, operands []*ast.Term, iter func(*ast.Term) error) error {
+		i, err := builtins.BigIntOperand(operands[0].Value, 1)
+		if err != nil {
+			return err
+		}
+		iOut, err := fn(i)
+		if err != nil {
+			return err
+		}
+		return iter(ast.NewTerm(builtins.IntToNumber(iOut)))
+	}
+}
+
+func builtinBitsArity2(fn bitsArity2) BuiltinFunc {
+	return func(_ BuiltinContext, operands []*ast.Term, iter func(*ast.Term) error) error {
+		i1, err := builtins.BigIntOperand(operands[0].Value, 1)
+		if err != nil {
+			return err
+		}
+		i2, err := builtins.BigIntOperand(operands[1].Value, 2)
+		if err != nil {
+			return err
+		}
+		iOut, err := fn(i1, i2)
+		if err != nil {
+			return err
+		}
+		return iter(ast.NewTerm(builtins.IntToNumber(iOut)))
+	}
+}
+
+func init() {
+	RegisterBuiltinFunc(ast.BitsOr.Name, builtinBitsArity2(bitsOr))
+	RegisterBuiltinFunc(ast.BitsAnd.Name, builtinBitsArity2(bitsAnd))
+	RegisterBuiltinFunc(ast.BitsNegate.Name, builtinBitsArity1(bitsNegate))
+	RegisterBuiltinFunc(ast.BitsXOr.Name, builtinBitsArity2(bitsXOr))
+	RegisterBuiltinFunc(ast.BitsShiftLeft.Name, builtinBitsArity2(bitsShiftLeft))
+	RegisterBuiltinFunc(ast.BitsShiftRight.Name, builtinBitsArity2(bitsShiftRight))
+}
diff --git a/topdown/bits_test.go b/topdown/bits_test.go
new file mode 100644
index 0000000000..809a7e2976
--- /dev/null
+++ b/topdown/bits_test.go
@@ -0,0 +1,148 @@
+// Copyright 2020 The OPA Authors.  All rights reserved.
+// Use of this source code is governed by an Apache2
+// license that can be found in the LICENSE file.
+
+package topdown
+
+import (
+	"fmt"
+	"math"
+	"testing"
+
+	"github.com/open-policy-agent/opa/ast"
+)
+
+func TestBuiltinBitsOr(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic bitwise-or", []string{`p[x] { x := bits.or(7, 9) }`}, `[15]`},
+		{"or with zero is value", []string{`p[x] { x := bits.or(50, 0) }`}, `[50]`},
+		{"lhs (float) error", []string{`p = x { x := bits.or(7.2, 42) }`}, &Error{Code: TypeErr, Message: "bits.or: operand 1 must be integer number but got floating-point number"}},
+		{
+			"rhs (wrong type-type) error",
+			[]string{`p = x { x := bits.or(7, "hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.or: invalid argument(s)")},
+		},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
+
+func TestBuiltinBitsAnd(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic bitwise-and", []string{`p[x] { x := bits.and(7, 9) }`}, `[1]`},
+		{"and with zero is and", []string{`p[x] { x := bits.and(50, 0) }`}, `[0]`},
+		{"lhs (float) error", []string{`p = x { x := bits.and(7.2, 42) }`}, &Error{Code: TypeErr, Message: "bits.and: operand 1 must be integer number but got floating-point number"}},
+		{
+			"rhs (wrong type-type) error",
+			[]string{`p = x { x := bits.and(7, "hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.and: invalid argument(s)")},
+		},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
+
+func TestBuiltinBitsNegate(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic bitwise-negate", []string{`p[x] { x := bits.negate(42) }`}, `[-43]`},
+		{"float error", []string{`p = x { x := bits.negate(7.2) }`}, &Error{Code: TypeErr, Message: "bits.negate: operand 1 must be integer number but got floating-point number"}},
+		{
+			"type error",
+			[]string{`p = x { x := bits.negate("hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.negate: invalid argument(s)")},
+		},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
+
+func TestBuiltinBitsXOr(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic bitwise-xor", []string{`p[x] { x := bits.xor(42, 3) }`}, `[41]`},
+		{"xor same is 0", []string{`p[x] { x := bits.xor(42, 42) }`}, `[0]`},
+		{"lhs (float) error", []string{`p = x { x := bits.xor(7.2, 42) }`}, &Error{Code: TypeErr, Message: "bits.xor: operand 1 must be integer number but got floating-point number"}},
+		{
+			"rhs (wrong type-type) error",
+			[]string{`p = x { x := bits.xor(7, "hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.xor: invalid argument(s)")},
+		},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
+
+func TestBuiltinBitsShiftLeft(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic shift-left", []string{`p[x] { x := bits.lsh(1, 3) }`}, `[8]`},
+		{"lhs (float) error", []string{`p = x { x := bits.lsh(7.2, 42) }`}, &Error{Code: TypeErr, Message: "bits.lsh: operand 1 must be integer number but got floating-point number"}},
+		{
+			"rhs (wrong type-type) error",
+			[]string{`p = x { x := bits.lsh(7, "hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.lsh: invalid argument(s)")},
+		},
+		{"rhs must be unsigned", []string{`p = x { x := bits.lsh(7, -1) }`}, &Error{Code: TypeErr, Message: "bits.lsh: operand 2 must be an unsigned integer number but got a negative integer"}},
+		{
+			"shift of max int32 doesn't overflow",
+			[]string{fmt.Sprintf(`p = x { x := bits.lsh(%d, 1) }`, math.MaxInt32)},
+			`4294967294`,
+		},
+		{
+			"shift of max int64 doesn't overflow, but it's lossy do to conversion to exponent type (see discussion in #2160)",
+			[]string{fmt.Sprintf(`p = x { x := bits.lsh(%d, 1) }`, math.MaxInt64)},
+			`18446744074000000000`,
+		},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
+
+func TestBuiltinBitsShiftRight(t *testing.T) {
+	tests := []struct {
+		note     string
+		rules    []string
+		expected interface{}
+	}{
+		{"basic shift-right", []string{`p[x] { x := bits.rsh(8, 3) }`}, `[1]`},
+		{"lhs (float) error", []string{`p = x { x := bits.rsh(7.2, 42) }`}, &Error{Code: TypeErr, Message: "bits.rsh: operand 1 must be integer number but got floating-point number"}},
+		{
+			"rhs (wrong type-type) error",
+			[]string{`p = x { x := bits.rsh(7, "hi") }`},
+			ast.Errors{ast.NewError(ast.TypeErr, nil, "bits.rsh: invalid argument(s)")},
+		},
+		{"rhs must be unsigned", []string{`p = x { x := bits.rsh(7, -1) }`}, &Error{Code: TypeErr, Message: "bits.rsh: operand 2 must be an unsigned integer number but got a negative integer"}},
+	}
+
+	for _, tc := range tests {
+		runTopDownTestCase(t, map[string]interface{}{}, tc.note, tc.rules, tc.expected)
+	}
+}
diff --git a/topdown/builtins/builtins.go b/topdown/builtins/builtins.go
index 415fcf7130..861167f398 100644
--- a/topdown/builtins/builtins.go
+++ b/topdown/builtins/builtins.go
@@ -92,6 +92,21 @@ func IntOperand(x ast.Value, pos int) (int, error) {
 	return i, nil
 }
 
+// BigIntOperand converts x to a big int. If the cast fails, a descriptive error
+// is returned.
+func BigIntOperand(x ast.Value, pos int) (*big.Int, error) {
+	n, err := NumberOperand(x, 1)
+	if err != nil {
+		return nil, NewOperandTypeErr(pos, x, "integer")
+	}
+	bi, err := NumberToInt(n)
+	if err != nil {
+		return nil, NewOperandErr(pos, "must be integer number but got floating-point number")
+	}
+
+	return bi, nil
+}
+
 // NumberOperand converts x to a number. If the cast fails, a descriptive error is
 // returned.
 func NumberOperand(x ast.Value, pos int) (ast.Number, error) {
@@ -159,8 +174,9 @@ func FloatToNumber(f *big.Float) ast.Number {
 // NumberToInt converts n to a big int.
 // If n cannot be converted to an big int, an error is returned.
 func NumberToInt(n ast.Number) (*big.Int, error) {
-	r, ok := new(big.Int).SetString(string(n), 10)
-	if !ok {
+	f := NumberToFloat(n)
+	r, accuracy := f.Int(nil)
+	if accuracy != big.Exact {
 		return nil, fmt.Errorf("illegal value")
 	}
 	return r, nil