diff --git a/src/cmd/cgo/main.go b/src/cmd/cgo/main.go
index 939e282ff0..77beb0992c 100644
--- a/src/cmd/cgo/main.go
+++ b/src/cmd/cgo/main.go
@@ -390,7 +390,7 @@ func main() {
 	// We already put _cgo_ at the beginning, so the main
 	// concern is other cgo wrappers for the same functions.
 	// Use the beginning of the 16 bytes hash of the input to disambiguate.
-	h := hash.New16()
+	h := hash.New32()
 	io.WriteString(h, *importPath)
 	var once sync.Once
 	var wg sync.WaitGroup
diff --git a/src/cmd/compile/internal/liveness/plive.go b/src/cmd/compile/internal/liveness/plive.go
index ac0c2dff0a..6c97858cf6 100644
--- a/src/cmd/compile/internal/liveness/plive.go
+++ b/src/cmd/compile/internal/liveness/plive.go
@@ -981,7 +981,7 @@ func (lv *Liveness) enableClobber() {
 		// Clobber only functions where the hash of the function name matches a pattern.
 		// Useful for binary searching for a miscompiled function.
 		hstr := ""
-		for _, b := range hash.Sum20([]byte(lv.f.Name)) {
+		for _, b := range hash.Sum32([]byte(lv.f.Name)) {
 			hstr += fmt.Sprintf("%08b", b)
 		}
 		if !strings.HasSuffix(hstr, h) {
diff --git a/src/cmd/compile/internal/types/fmt.go b/src/cmd/compile/internal/types/fmt.go
index 0dba510ac4..139defafe2 100644
--- a/src/cmd/compile/internal/types/fmt.go
+++ b/src/cmd/compile/internal/types/fmt.go
@@ -646,7 +646,7 @@ func SplitVargenSuffix(name string) (base, suffix string) {
 func TypeHash(t *Type) uint32 {
 	p := t.LinkString()
 
-	// Using 16 bytes hash is overkill, but reduces accidental collisions.
-	h := hash.Sum16([]byte(p))
+	// Using a cryptographic hash is overkill but minimizes accidental collisions.
+	h := hash.Sum32([]byte(p))
 	return binary.LittleEndian.Uint32(h[:4])
 }
diff --git a/src/cmd/internal/hash/hash.go b/src/cmd/internal/hash/hash.go
index a37368f50e..cdb24a2645 100644
--- a/src/cmd/internal/hash/hash.go
+++ b/src/cmd/internal/hash/hash.go
@@ -5,69 +5,26 @@
 // Package hash implements hash functions used in the compiler toolchain.
 package hash
 
-// TODO(rsc): Delete the 16 and 20 forms and use 32 at all call sites.
-
 import (
 	"crypto/sha256"
 	"hash"
 )
 
-const (
-	// Size32 is the size of the 32-byte hash checksum.
-	Size32 = 32
-	// Size20 is the size of the 20-byte hash checksum.
-	Size20 = 20
-	// Size16 is the size of the 16-byte hash checksum.
-	Size16 = 16
-)
+// Size32 is the size of the 32-byte hash functions [New32] and [Sum32].
+const Size32 = 32
 
-type shortHash struct {
-	hash.Hash
-	n int
-}
-
-func (h *shortHash) Sum(b []byte) []byte {
-	old := b
-	sum := h.Hash.Sum(b)
-	return sum[:len(old)+h.n]
-}
-
-// New32 returns a new [hash.Hash] computing the 32 bytes hash checksum.
+// New32 returns a new [hash.Hash] computing the 32-byte hash checksum.
+// Note that New32 and [Sum32] compute different hashes.
 func New32() hash.Hash {
 	h := sha256.New()
 	_, _ = h.Write([]byte{1}) // make this hash different from sha256
 	return h
 }
 
-// New20 returns a new [hash.Hash] computing the 20 bytes hash checksum.
-func New20() hash.Hash {
-	return &shortHash{New32(), 20}
-}
-
-// New16 returns a new [hash.Hash] computing the 16 bytes hash checksum.
-func New16() hash.Hash {
-	return &shortHash{New32(), 16}
-}
-
-// Sum32 returns the 32 bytes checksum of the data.
-func Sum32(data []byte) [Size32]byte {
+// Sum32 returns a 32-byte checksum of the data.
+// Note that Sum32 and [New32] compute different hashes.
+func Sum32(data []byte) [32]byte {
 	sum := sha256.Sum256(data)
-	sum[0] ^= 1 // make this hash different from sha256
+	sum[0] ^= 0xff // make this hash different from sha256
 	return sum
 }
-
-// Sum20 returns the 20 bytes checksum of the data.
-func Sum20(data []byte) [Size20]byte {
-	sum := Sum32(data)
-	var short [Size20]byte
-	copy(short[:], sum[4:])
-	return short
-}
-
-// Sum16 returns the 16 bytes checksum of the data.
-func Sum16(data []byte) [Size16]byte {
-	sum := Sum32(data)
-	var short [Size16]byte
-	copy(short[:], sum[8:])
-	return short
-}
diff --git a/src/cmd/internal/obj/objfile.go b/src/cmd/internal/obj/objfile.go
index bc22765abc..3299fbf4e6 100644
--- a/src/cmd/internal/obj/objfile.go
+++ b/src/cmd/internal/obj/objfile.go
@@ -494,7 +494,7 @@ func contentHash64(s *LSym) goobj.Hash64Type {
 // For now, we assume there is no circular dependencies among
 // hashed symbols.
 func (w *writer) contentHash(s *LSym) goobj.HashType {
-	h := hash.New20()
+	h := hash.New32()
 	var tmp [14]byte
 
 	// Include the size of the symbol in the hash.
diff --git a/src/cmd/internal/obj/sym.go b/src/cmd/internal/obj/sym.go
index 8872579050..08c50ec72b 100644
--- a/src/cmd/internal/obj/sym.go
+++ b/src/cmd/internal/obj/sym.go
@@ -216,7 +216,7 @@ func (ctxt *Link) Int128Sym(hi, lo int64) *LSym {
 
 // GCLocalsSym generates a content-addressable sym containing data.
 func (ctxt *Link) GCLocalsSym(data []byte) *LSym {
-	sum := hash.Sum16(data)
+	sum := hash.Sum32(data)
 	str := base64.StdEncoding.EncodeToString(sum[:16])
 	return ctxt.LookupInit(fmt.Sprintf("gclocals·%s", str), func(lsym *LSym) {
 		lsym.P = data
diff --git a/src/cmd/link/internal/ld/elf.go b/src/cmd/link/internal/ld/elf.go
index e6a525198f..6ff1d94383 100644
--- a/src/cmd/link/internal/ld/elf.go
+++ b/src/cmd/link/internal/ld/elf.go
@@ -1690,11 +1690,11 @@ func (ctxt *Link) doelf() {
 		sb.SetType(sym.SRODATA)
 		ldr.SetAttrSpecial(s, true)
 		sb.SetReachable(true)
-		sb.SetSize(hash.Size20)
+		sb.SetSize(hash.Size32)
 		slices.SortFunc(ctxt.Library, func(a, b *sym.Library) int {
 			return strings.Compare(a.Pkg, b.Pkg)
 		})
-		h := hash.New20()
+		h := hash.New32()
 		for _, l := range ctxt.Library {
 			h.Write(l.Fingerprint[:])
 		}
diff --git a/src/cmd/link/internal/ld/lib.go b/src/cmd/link/internal/ld/lib.go
index 2d8f964f35..b114ca2a3d 100644
--- a/src/cmd/link/internal/ld/lib.go
+++ b/src/cmd/link/internal/ld/lib.go
@@ -1022,7 +1022,7 @@ func typeSymbolMangle(name string) string {
 		return name
 	}
 	if isType {
-		hb := hash.Sum20([]byte(name[5:]))
+		hb := hash.Sum32([]byte(name[5:]))
 		prefix := "type:"
 		if name[5] == '.' {
 			prefix = "type:."
@@ -1035,7 +1035,7 @@ func typeSymbolMangle(name string) string {
 	if j == -1 || j <= i {
 		j = len(name)
 	}
-	hb := hash.Sum20([]byte(name[i+1 : j]))
+	hb := hash.Sum32([]byte(name[i+1 : j]))
 	return name[:i+1] + base64.StdEncoding.EncodeToString(hb[:6]) + name[j:]
 }
 
diff --git a/src/cmd/objdump/objdump_test.go b/src/cmd/objdump/objdump_test.go
index 0f3a183c61..0d6f608a3f 100644
--- a/src/cmd/objdump/objdump_test.go
+++ b/src/cmd/objdump/objdump_test.go
@@ -134,9 +134,9 @@ func testDisasm(t *testing.T, srcfname string, printCode bool, printGnuAsm bool,
 		goarch = f[1]
 	}
 
-	hash := hash.Sum16([]byte(fmt.Sprintf("%v-%v-%v-%v", srcfname, flags, printCode, printGnuAsm)))
+	hash := hash.Sum32([]byte(fmt.Sprintf("%v-%v-%v-%v", srcfname, flags, printCode, printGnuAsm)))
 	tmp := t.TempDir()
-	hello := filepath.Join(tmp, fmt.Sprintf("hello-%x.exe", hash))
+	hello := filepath.Join(tmp, fmt.Sprintf("hello-%x.exe", hash[:16]))
 	args := []string{"build", "-o", hello}
 	args = append(args, flags...)
 	args = append(args, srcfname)